)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"835f2996195389936d9d2ee6052af01719a7ec0a","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":2,"id":"fadef5d2_429771b2","updated":"2026-10-16 13:49:03.000000000","message":"recheck\n\n\nbuild gone","commit_id":"6cb6207ee6cad22518f615a9cabcbd893115941a"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"446872b9593fe69e742aa9a46c53cca9d331d40b","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"da239190_230b1151","updated":"2026-10-20 17:30:03.000000000","message":"Adding a 1-week timebox here for Roll Call Votes from the TC. Please use your Roll-Call +1 or -1 votes to indicate your position here. Please use Roll-Call 0 for abstain \n\nIt\u0027d be really helpful to have comments though because objective positions may not show why you\u0027re supportive, not supportive or ambivalent","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"0fb4c220247e61877a0e3a9c7f2ccb4b5376371f","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"7a1ae5ef_3dfb2453","updated":"2026-10-20 11:11:55.000000000","message":"CRA requirements just landed on my table as responsible for IAM at my current employer. As such adding some concrete findings/requirements from this side as well into the resolution. While CRA does not directly affects upstream (steward), it affects every downstream fully. Keystone is a \"class I - important\" application according to the CRA (same classification likely apply for Neutron and maybe Barbican, while other services does not seem to get any classification at all)","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":15993,"name":"Amy Marrich","display_name":"Amy Marrich (spotz)","email":"amy@demarco.com","username":"amarrich"},"change_message_id":"9c9c66eb31067b9d076cef9cc10581c0d7f0f320","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"0310db6d_09740508","updated":"2026-10-22 16:23:32.000000000","message":"For right now I\u0027m going to Abstain on this. I am not against the idea of bringing in new languages to the project, though I hold to the thought that any existing project being rewritten in a new language should be a new project. Base on Dan\u0027s concerns in the ticket and fungi during the meeting I would like to see those addressed before I would +1","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"c2848624_629581c2","updated":"2026-10-30 06:32:06.000000000","message":"Hey Artem,\n\nApologies for the delay to point out the minutiae. In fact, most of the points I wanted to make were made by others and I offered support. In my view, as I said earlier with my top level comment, I\u0027m not opposed to Rust per se. Many arguments in this are opinions and not facts; I want to encourage us to explore Rust and what bringing it to the community would look like. I don\u0027t really think the heavy process we set up is needed before trying to get one or two deliverables fleshed out by going through the pain of introducing the PTI, CI infrastructure, iterating and building community expertise.\n\nWhile this happens, I\u0027m totally for an experiment bringing a Rust-based IAM alongside Python Keystone. You\u0027ve articulated many improvements that operators have lauded on the mailing list thread. However, adopting something like this will take time and we should accept that reality. In that reality, I am not for dropping our investment in the Python-based Keystone. We must allow bugfixes and features to land, and for Keystone to continue to improve.\n\nThe small Keystone maintainer team can\u0027t do it all though, and therein lies our biggest challenge. In my inline comments, I\u0027ve asked for some changes and clarifications. I\u0027m -1 until those are sufficiently addressed.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"33371adbf939dc93dcc9b5a1d48427bbf8bd51ea","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"82a596dc_b596ce5b","updated":"2026-10-21 16:50:33.000000000","message":"I could rewrite, or actually change the scope of the resolution, to start allowing rust to be embedded into python to start building experience if that helps anybody. I will not do this when nobody think it would address their concerns. It would be great to allow OpenStack evolve without arbitrary project team being hardly limited by rules written without considering their unique demands.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8c1a8e3105139bac19d3b56c88bd0adc606b0d1d","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"6af67583_c7c9d62f","updated":"2026-10-27 15:54:02.000000000","message":"I strongly believe that our community should be open to experimentation and innovation. If there are meaningful gains from using a particular approach, platform, or programming language, I’m glad to explore that route.\n\nA new language brings substantial requirements on the community. Defining the Project Testing Interface (PTI), CI jobs, packaging, and delivery and staffing infra, QA, security and release teams - will be challenging - but we can\u0027t expect all of those to be fully fleshed out prior to beginning the experiment - we have to start somewhere. We added Golang support years ago, but we struggled to push it forward because it is a heavy mission to embark on. I fear this proposal might run into the same problem. To boot, currently, the proposal relies on a very small subset of invested leaders (until recent days, just Artem).\n\nAccepting this does not mean that Keystone’s Rust rewrite plan is automatically accepted, as Artem also acknowledges. I share the opinion that a rewrite is much better pitched as a parallel service with a distinct name, and I would expect any critical service replacement to undergo substantial soak time prior to calling it production-ready.\n\nIt is true that this proposal would be far more palatable if it were pitching a new tool, library, or service that is naturally better suited for Rust than Python. I previously tried making the argument for onboarding a Rust-based SDK or CLI first, as a way to grind through the hard work of defining the PTI, CI workflows, packaging, and operational procedures.\n\nAs others have pointed out, there are also several inaccuracies in the current proposal that will need to be addressed before we can move forward. I’ve been away recently and need access to internet that isn\u0027t as patchy before I can dive in and leave meaningful inline comments, but I wanted to drop these high-level thoughts here in the meantime.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"f33c2f66dc550f2439425049ddca89f8cc5664f4","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"f3b61351_bad1acb3","updated":"2026-10-21 06:59:29.000000000","message":"I would address comments in the resolution when there is a real interest to have it improved or questions answered","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"2a49234304a9a4a28142c28e1d935d9bb49afa95","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"cada5843_962d2991","updated":"2026-10-20 18:02:05.000000000","message":"I\u0027m adding some comments here, shamelessly admitting they\u0027re not a full explanation of why I don\u0027t want to do this (somewhat because I hit on more of those just late last week). I wasn\u0027t in favor of Go at the time (even with the massively smaller scope it was proposing) and I\u0027m similarly opposed to this. It has nothing to do with liking Rust or thinking Rust is good or bad in isolation.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":22629,"name":"Michal Nasiadka","email":"mnasiadka@gmail.com","username":"mnasiadka"},"change_message_id":"35b99eed92fc0ec64a32f6159ddba5341d413d97","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":3,"id":"db9cf262_aa67358b","updated":"2026-10-26 11:11:31.000000000","message":"I\u0027m voting -1, and I want to explain why, because it has nothing to do with Rust as a language.\nSome arguments here are fair, for example on authorization-logic correctness. The new capabilities (native OIDC OP, WebAuthn, SCIM) are real operator asks, and I\u0027d like to see them delivered.\nMy concern is who would carry this, and whether the community has the machinery and people to support it.\n\n1. Keystone team capacity\n\nOver the last 12 months, openstack/keystone merged about 234 changes. Almost all approvals came from three people; the remaining approvals were mostly stable and release maintenance.\nPython Keystone can\u0027t go away. Every deployment, every other service\u0027s keystonemiddleware, and all the stable branches depend on it.\nAdding a Rust component means this same small group would maintain two implementations and the compatibility between them. On top of that, they\u0027d take on CI, packaging, release and stable work that other teams handle for Python today. Christian\u0027s examples (the TOTP test race, Flask-RESTful) show that small fixes in the existing codebase already go unaddressed. I don\u0027t see how a second codebase improves that. The \"Community cost\" section covers all of this in two sentences, and I don\u0027t think it reflects the real cost.\n\n2. No release machinery for Rust deliverables\n\nOur release tooling and processes are built around Python sdists and wheels.\nOpenStack never built a real nodejs release process for Skyline. It avoided the problem by wrapping the JS build output in a Python package. A standalone Rust service can\u0027t do that.\nFor Rust, nothing exists yet. We\u0027d need:\n- crate mirroring and caching in the gate\n- a decision on what we publish (source tarballs, crates.io, binaries?)\n- a policy on which Rust compiler versions we support, aligned with what distributions ship. Debian and Ubuntu build from source with their frozen distro rustc, so a service (or any of its transitive crates) that needs a newer compiler can\u0027t be packaged without patching, vendoring older crates, or backporting a compiler\n- a policy for vendored and locked dependencies on stable branches\n- a process for re-releasing stable branches when a locked dependency gets a security fix. Unlike Python, where distributions upgrade a shared library on their own, crates compiled into a binary need a new release from us\n\n3. Vulnerability management\n\nKeystone is the most security-sensitive service we ship. Its advisories need VMT members who can triage reports, judge impact, and review embargoed fixes quickly. I\u0027m not aware of anyone on the VMT with significant Rust experience. If only the author can properly assess security reports for this code, we have a single point of failure in exactly the place we can least afford one.\n\n4. OpenDev and TaCT SIG impact\n\nAs a TaCT SIG co-chair, I\u0027d add the CI side. Hosting an unofficial project on OpenDev is fine, but making it an official, gating deliverable is a different matter. We\u0027d need a crates.io mirror or proxy (or vendored crates) so jobs don\u0027t fetch hundreds of crates from the internet, Rust toolchains on the test nodes that match whatever compiler policy we choose, and build caching or prebuilt artifacts so that devstack-based jobs don\u0027t each compile the service from scratch. On the OpenStack side there\u0027s no Rust PTI, no job templates, and no devstack integration, and the same small group of people would have to review and debug all of it on stable branches.\n\nI agree with Amy, Dan and Dmitriy that this should be a new deliverable, ideally with its own project team, and not a replacement for Keystone under the current team. Christian\u0027s proposal is a reasonable starting point.\n\nBefore I could support a Step 1 approval, I\u0027d want to see:\n- named people from named organisations committing to maintain the Rust service, its CI, releases, stable branches and security handling, beyond the current Keystone cores\n- someone with Rust experience committing to work with the release team and the VMT, ideally by joining them\n- contributors committing to implement and maintain the Rust CI tooling (mirroring, toolchains, job templates, PTI, devstack integration) together with the OpenDev admins and TaCT SIG, instead of assuming existing volunteers will absorb it\n\nMoving the code to OpenDev now as an unofficial repository doesn\u0027t need this resolution. I\u0027d encourage it either way, because it would let us see how the project is staffed and reviewed in practice before we vote on making it official.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":167,"name":"Christian Berendt","email":"berendt@osism.tech","username":"berendt"},"change_message_id":"efdce0edf4fd4d8b6ee023a53e85270cd19b7c3a","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"654f4383_6b4b17cc","updated":"2026-10-26 08:05:24.000000000","message":"I\u0027ve spent some time over the last days looking at Keystone, partly in the\ncontext of CobaltCore. Two small examples: the TOTP tests fail intermittently\nbecause of a clock race [1], and Flask-RESTful, abandoned for years, is still\na runtime dependency [2]. Both fixes were small. That nobody had done them is,\nto me, the actual data point: Keystone\u0027s problem today is not the language,\nit is that very few people are working on it. Downstream this shows up as a\nCI that is regularly red, a supply chain that is hard to defend in audits, and\nfeatures we need (FIPS mode, SCIM) that are on nobody\u0027s roadmap.\n\nI agree with several of the review comments. Parts of the technical case are\noutdated (as Dan pointed out, eventlet is no longer a Keystone dependency, and\nfree-threaded Python changes the GIL discussion), and the community cost\nsection needs to be much more than two sentences. I also agree with Amy, Dan\nand Dmitriy that this should be a new deliverable and not a rewrite that\nreplaces Keystone under the existing team. Artem already said he is open to\nthat.\n\nOn the CRA: Dan is right that its obligations fall on manufacturers and\ndistributors, not on us as upstream. Still, it would help adoption a lot if we\nkept these requirements in view, because what we decide upstream determines\nhow hard they are to meet downstream. Given how much demand for this there\nappears to be, especially from European operators, I think that is worth doing\neven without a formal obligation.\n\nWhat I don\u0027t think helps us is a plain no. The Rust implementation exists,\nArtem will continue it either way, and the demand goes beyond one or two\norganizations. If we reject the use case, the work continues outside of\nOpenDev, where we have neither visibility nor influence, and the Python\nKeystone does not gain a single contributor. There is no plan B in this\nreview.\n\nRust is also no longer foreign to our stack: pyca/cryptography already needs a\nRust toolchain, Debian and other distributions ship one, the kernel uses it,\nand zigo is already working with Artem on packaging.\n\nSo I\u0027d like to propose a path instead of a verdict:\n\n1. Treat it as a new deliverable (a new IAM service, possibly its own project\n   team) under its own name, not as a replacement for Keystone. Python Keystone\n   stays as it is; API compatibility is the goal.\n2. Move the code from GitHub to OpenDev now as an unofficial repository. As\n   fungi and clarkb said in #openstack-tc, the language policy restricts\n   official releases, not experiments on OpenDev, and that is where we get\n   shared CI, reviews and visibility.\n3. Narrow this resolution to the use case of this one service, with the Step 2\n   requirements (CI, source releases, a dependency and rebuild policy, stable\n   and VMT handling) as explicit preconditions before anything is released.\n   No ecosystem-wide Rust commitment and no oslo equivalent up front.\n4. Rewrite the community cost section with named commitments: who maintains\n   CI, releases, stable branches and security handling, and which\n   organizations staff it.\n5. Drop the arguments that don\u0027t hold and keep the ones that do: authorization\n   logic correctness, a native OIDC provider, WebAuthn, SBOM.\n\nI\u0027m voting +1 to support the effort itself. My +1 is for the use case, the\n\"why\", not for every argument in the current text, and I\u0027d expect a revised\npatch set along the lines above. Goutham, I\u0027d suggest we don\u0027t close the roll\ncall and give Artem time for that revision. It would also give all of us time\nto think about whether a path like the one above is an option we could support,\nrather than deciding on the current text alone.\n\n[1] https://review.opendev.org/c/openstack/keystone/+/1012064\n[2] https://review.opendev.org/c/openstack/keystone/+/1011990","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"c4ddf2254c5bf9f1179bda429f09a9b11e1de548","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"3dad61d5_2b1aaa65","updated":"2026-10-22 17:23:30.000000000","message":"and one last point for today. It is not constructive when people not knowing the technology are trying to argument whether it is better or not. Question which OSSA would have been prevented is also very speculative. Different languages use different concepts (they namely exist for that reason), so just because python and rust are both memory safe doesn\u0027t say in Rust you would actually had or not had this particular vulnerability, because the language guides you to treat data differently. We had and still have plenty of vulnerabilities related to data reuse and race between requests/threads. In Rust this in principally not possible due to the unique data ownership model. Just stating both languages are memory safe is not a fair comparison, it doesn\u0027t map 1to1.\nIAM is where any compiled language is better than dynamic just by the definition, and Rust is better than Go because it was designed to prevent certain classes of typical errors. IAM requires a very high throughput with a very low latency. Other OpenStack services do not have this concern to the same degree. Not every OpenStack installation is affected by that, but big public clouds do permanently raise this concern.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":13252,"name":"Dr. Jens Harbott","display_name":"Jens Harbott (frickler)","email":"frickler@offenerstapel.de","username":"jrosenboom"},"change_message_id":"147962028c763e89ce0e46a3c558bc94dd776433","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":3,"id":"201cf53b_e311da78","updated":"2026-11-02 11:09:50.000000000","message":"one more note: I think it will be helpful to make a clearer distinction between keystone the deliverable (and the corresponding openstack/keystone repo) and the keystone project team as defined in the governance.\n\nI assume it is pretty clear that a new \"identity service written in rust\" deliverable would be hosted in a new git repo. (Or is that wrong already?)\n\nIt is less clear to me whether it could and should re-use the keystone name/branding or whether a different naming would be better to avoid confusing deployers, while we also keep supporting the current keystone implementation for the foreseeable future.\n\nIt is also not clear to me whether the new deliverable could and should belong to the same current project team, or whether a dedicated team definition would be better.\n\nNone of this needs to be decided right away, but since some of these questions are referenced by your reasoning in this resolution, I think some clarification regarding your ideas about this could be helpful.\n\nIn general I support going ahead with this proposal.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"60b17c291198930ff4ba4eaa9458d8c04d660641","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"9e5b4d3b_d0c0f6f4","in_reply_to":"0310db6d_09740508","updated":"2026-10-22 16:30:56.000000000","message":"\u003e I hold to the thought that any existing project being rewritten in a new language should be a new project.\n\nI feel the same way, FWIW.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"c0d106b0e1a38a4c51500622a578c85873a1ed01","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"a8095c87_e5e23e62","in_reply_to":"0310db6d_09740508","updated":"2026-10-22 16:38:39.000000000","message":"Thanks Amy for trying to be constructive. I think I have shown over the last years readiness to adapt, also with regards to this resolution whith roots started 2 years back.\n\nWe can easily talk about it being a new service, for sure. There would be challenges to solve caused exactly by that, but those are resolvable. Resolution on itself does not say we rewrite Keystone in Rust, but rather Keystone, and here I would reposition it as IAM service of OpenStack, has unique challenges which are much more efficiently addressable in Rust. IAM is a very sensitive part and has different requirements and expectations from the rest of OpenStack.\nWrt delivery model: on one side I am working with zigo to address packaging. But also we have possibility to package even the binary as python wheel (or some other ways to deliver rust code to the operator to solve the packaging).\n\nAs such I would again like to repeat that we can discuss all of this constructively. Resolution is just a door for us to start talking \"how\".","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"975a3f77fce84f9142af90fa0f166d159e1468cb","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":3,"id":"beae09cc_27e4b0fa","in_reply_to":"11903379_842cf290","updated":"2026-10-30 06:49:18.000000000","message":"Yes, the process is also open for debate.. In this case I see several TC members coming to the same conclusion; so maybe it\u0027s productive for us to evolve the process. We don\u0027t have to hold it up for that though. Let\u0027s as a team attempt to define what the process should be with this proposal.\n\nGolang failed here because it never got to Step 2. \n\nWe don\u0027t even have to violate it in entirety imo. Sketch to us what the step 2 would look like because you\u0027re getting these questions; or I do see value in Christian\u0027s path of moving things to opendev to showcase that integration challenges can be met, and these questions can be answered with that experience. If this experiment succeeds, step 1 sorta becomes a formality.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"313213e3de9f028fa6df8a1ed51ba72591c254a9","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"11903379_842cf290","in_reply_to":"654f4383_6b4b17cc","updated":"2026-10-26 10:28:33.000000000","message":"Thank you very much Christian. Just a heads-up, I was presenting it already a year ago during the Paris Summit - SCIM is done in the Rust implementation. It was actually one of the first things I started the whole project for (starting as building an extension to Keystone). Back those days the context was much smaller, but heavily evolved since. FIPS support in Keystone is part of Rust project now as well, as well as technical solutions to those of us falling under much stricter audit requirements than maybe others. Requirements to public cloud are not the same as to private cloud, we can\u0027t ignore them if we try to care.\n\nBTW: on the conference I participated during the weekend (with an openinfra track) one answer during the Q\u0026A made be rolling from laugh (https://github.com/ironcore-dev/ironcore): \"Q: why did you chose this approach over X [perhaps Metal, but I forgot the exact project name]? A: Because it is based on Ironic and we didn\u0027t want to run this heavy python app\" \u003d\u003e So with the intention to prevent a community split the absolute opposite was already achieved. [I didn\u0027t manufactured this, it happened for real]. I think we should be addressing technical items related to having multiple programming languages in the cloud ecosystem, rather than simply saying we only allow one language: \"don\u0027t like it? nobody holds you here\". Nearly every developer today submits patches produced with the help of AI without having a deep (and often even light) knowledge of the project.\n\nI can address points mentioned in the resolution, but I would like to remind, that step 1 of the process for adoption of the new programming language is to present the usecase and not to agree on the technical implementation. We shouldn\u0027t jump in front of the train and instead try to answer questions in the sequence. The more context we try to mix together the harder it gets. But if people think in this case it is necessary to violate (or let\u0027s name it \"re-interpret\") the process to come to the conclusion I am fine with it. To be honest, I myself do not strictly agree with how the process is described, it mandates certain things which might conceptually simply not apply to the different programming languages, so maybe the whole process needs to be revisited.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"a84343eaf11d7425f6b0a467957f058a0742a098","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":3,"id":"f2c42630_2e57bce4","in_reply_to":"a8095c87_e5e23e62","updated":"2026-10-22 16:51:38.000000000","message":"Dan, thanks as well, it took a while to write answer to Amy from mobile phone, that I missed your response. Resolution does not in any way talk about rewriting Keystone in Rust, but provides usecase as expected by the current process. Remember that I was asking you to treat the resolution like from some arbitrary service stating why it would help us solve some our problems. It is \"why\", and not the \"how\". I really want us not to repeat mistakes of the past rejecting other project and ideas just because \"you are not like us\" or \"I do not know this technology\".\n\nSo if you have concrete points that would flip you vote please let me know in detail. I just don\u0027t want to start rewriting the resolution to confuse others if that is not going to change your mind.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"9c57ec96cae4011938f7ca3678c8eb1faf1461c9","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":3,"id":"785d3c0d_153cfc49","in_reply_to":"db9cf262_aa67358b","updated":"2026-10-26 11:35:16.000000000","message":"Thanks for the response. \n\n\u003e No release machinery for Rust deliverables\n\nAs I explain above there are ways to package Rust binaries into python wheels, we could use library approach, etc.\n\nOverall you require step 2 to be **fully** completed before talking about step 1 turning the current official process head down. This is not a constructive compromise from my pov.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":4,"id":"b7532429_72141681","updated":"2026-11-06 16:31:04.000000000","message":"Due to the major rewrite of arguments I have marked all comments as done. Either they are addressed directly or the lines are gone completely. Feel free to repost comments if you don\u0027t feel they are addressed.","commit_id":"0fe26ceb5986911f1ceb2d9db94552a19099f9c2"},{"author":{"_account_id":14250,"name":"Grzegorz Grasza","email":"xek@redhat.com","username":"xek"},"change_message_id":"2206d8b3c01266ab740cacc228fccfc1f5d6b032","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":5,"id":"0901bdbd_59a50d0a","updated":"2026-11-13 14:32:46.000000000","message":"I support allowing this effort to proceed. My main concern is that we might block innovation just because we can’t be sure it will succeed.\n\nKeystone needs more than incremental performance improvements. I would like OpenStack to have room to explore simpler, standards-based identity architectures, with less maintenance and a smaller attack surface. There is value in better interoperability with other identity systems and using established standards instead of Keystone-specific mechanisms where suitable alternatives exist.\n\nThat does not establish that Rust is necessary, or that Python Keystone cannot evolve in the same direction. But we do not yet know which approach will best deliver those outcomes, and I would prefer to learn from practical experience rather than settle that question through predictions alone.","commit_id":"4c063c7f72a694d866e511095b71a57c6e2e3cf3"},{"author":{"_account_id":15993,"name":"Amy Marrich","display_name":"Amy Marrich (spotz)","email":"amy@demarco.com","username":"amarrich"},"change_message_id":"a6af51bb85461c5658c9e1c2264088d796fcf3e3","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":5,"id":"65c7277d_0bdb776a","updated":"2026-11-13 21:54:32.000000000","message":"I\u0027m still hanging out as a 0 due to the many mentions of keystone, though I do realize some are used as a comparison to say it\u0027s faster. I would just really like this to stand on it\u0027s own merit","commit_id":"4c063c7f72a694d866e511095b71a57c6e2e3cf3"}],"resolutions/20260724-rust-use-case.rst":[{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":8,"context_line":"Following the process outlined in the Requirements for language additions to"},{"line_number":9,"context_line":"the OpenStack Ecosystem [2]_, this resolution presents the formal use-case analysis"},{"line_number":10,"context_line":"(Step 1) for adding Rust as an officially supported language under the"},{"line_number":11,"context_line":"OpenStack governance model. The primary catalyst for this proposal is the"},{"line_number":12,"context_line":"critical need to modernize OpenStack\u0027s core identity infrastructure (Keystone)"},{"line_number":13,"context_line":"and adjacent ecosystem projects. This modernization is driven by six"},{"line_number":14,"context_line":"converging pressures: mounting regulatory expectations around memory safety in"},{"line_number":15,"context_line":"critical infrastructure, the need for parallelism in CPU-bound cryptographic"},{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"}],"source_content_type":"text/x-rst","patch_set":3,"id":"d014eaba_3be19c0b","line":13,"range":{"start_line":11,"start_character":28,"end_line":13,"end_character":32},"updated":"2026-10-20 17:50:14.000000000","message":"After reading the document I still fail to see how modernization is requiring a new language in the ecosystem.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":8,"context_line":"Following the process outlined in the Requirements for language additions to"},{"line_number":9,"context_line":"the OpenStack Ecosystem [2]_, this resolution presents the formal use-case analysis"},{"line_number":10,"context_line":"(Step 1) for adding Rust as an officially supported language under the"},{"line_number":11,"context_line":"OpenStack governance model. The primary catalyst for this proposal is the"},{"line_number":12,"context_line":"critical need to modernize OpenStack\u0027s core identity infrastructure (Keystone)"},{"line_number":13,"context_line":"and adjacent ecosystem projects. This modernization is driven by six"},{"line_number":14,"context_line":"converging pressures: mounting regulatory expectations around memory safety in"},{"line_number":15,"context_line":"critical infrastructure, the need for parallelism in CPU-bound cryptographic"},{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"}],"source_content_type":"text/x-rst","patch_set":3,"id":"b5811681_266d20e3","line":13,"range":{"start_line":11,"start_character":28,"end_line":13,"end_character":32},"in_reply_to":"d014eaba_3be19c0b","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":11,"context_line":"OpenStack governance model. The primary catalyst for this proposal is the"},{"line_number":12,"context_line":"critical need to modernize OpenStack\u0027s core identity infrastructure (Keystone)"},{"line_number":13,"context_line":"and adjacent ecosystem projects. This modernization is driven by six"},{"line_number":14,"context_line":"converging pressures: mounting regulatory expectations around memory safety in"},{"line_number":15,"context_line":"critical infrastructure, the need for parallelism in CPU-bound cryptographic"},{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"},{"line_number":17,"context_line":"for embedded distributed consensus, a recurring pattern of authorization-logic"}],"source_content_type":"text/x-rst","patch_set":3,"id":"55835480_2000803a","line":14,"range":{"start_line":14,"start_character":55,"end_line":14,"end_character":75},"updated":"2026-10-20 17:50:14.000000000","message":"Python is memory safe.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":11,"context_line":"OpenStack governance model. The primary catalyst for this proposal is the"},{"line_number":12,"context_line":"critical need to modernize OpenStack\u0027s core identity infrastructure (Keystone)"},{"line_number":13,"context_line":"and adjacent ecosystem projects. This modernization is driven by six"},{"line_number":14,"context_line":"converging pressures: mounting regulatory expectations around memory safety in"},{"line_number":15,"context_line":"critical infrastructure, the need for parallelism in CPU-bound cryptographic"},{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"},{"line_number":17,"context_line":"for embedded distributed consensus, a recurring pattern of authorization-logic"}],"source_content_type":"text/x-rst","patch_set":3,"id":"b39d2ed4_23b70044","line":14,"range":{"start_line":14,"start_character":55,"end_line":14,"end_character":75},"in_reply_to":"55835480_2000803a","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":12,"context_line":"critical need to modernize OpenStack\u0027s core identity infrastructure (Keystone)"},{"line_number":13,"context_line":"and adjacent ecosystem projects. This modernization is driven by six"},{"line_number":14,"context_line":"converging pressures: mounting regulatory expectations around memory safety in"},{"line_number":15,"context_line":"critical infrastructure, the need for parallelism in CPU-bound cryptographic"},{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"},{"line_number":17,"context_line":"for embedded distributed consensus, a recurring pattern of authorization-logic"},{"line_number":18,"context_line":"and concurrency-safety bugs in the current implementation, and the need to"}],"source_content_type":"text/x-rst","patch_set":3,"id":"7897c964_f4aeed43","line":15,"range":{"start_line":15,"start_character":34,"end_line":15,"end_character":62},"updated":"2026-10-20 17:50:14.000000000","message":"Python has no-GIL mainstreamed now","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":12,"context_line":"critical need to modernize OpenStack\u0027s core identity infrastructure (Keystone)"},{"line_number":13,"context_line":"and adjacent ecosystem projects. This modernization is driven by six"},{"line_number":14,"context_line":"converging pressures: mounting regulatory expectations around memory safety in"},{"line_number":15,"context_line":"critical infrastructure, the need for parallelism in CPU-bound cryptographic"},{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"},{"line_number":17,"context_line":"for embedded distributed consensus, a recurring pattern of authorization-logic"},{"line_number":18,"context_line":"and concurrency-safety bugs in the current implementation, and the need to"}],"source_content_type":"text/x-rst","patch_set":3,"id":"a1dc0dc9_1d786ae5","line":15,"range":{"start_line":15,"start_character":34,"end_line":15,"end_character":62},"in_reply_to":"7897c964_f4aeed43","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":15,"context_line":"critical infrastructure, the need for parallelism in CPU-bound cryptographic"},{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"},{"line_number":17,"context_line":"for embedded distributed consensus, a recurring pattern of authorization-logic"},{"line_number":18,"context_line":"and concurrency-safety bugs in the current implementation, and the need to"},{"line_number":19,"context_line":"build modern, crypto-heavy authentication capabilities - a native OAuth2/OIDC"},{"line_number":20,"context_line":"provider and WebAuthn/passkey support - that Keystone lacks today and must"},{"line_number":21,"context_line":"build regardless of language."}],"source_content_type":"text/x-rst","patch_set":3,"id":"a59aafc9_ff42595a","line":18,"range":{"start_line":18,"start_character":4,"end_line":18,"end_character":57},"updated":"2026-10-20 17:50:14.000000000","message":"concurency safety derives not from the language, but from the implementation, I agree here. But it doesn\u0027t justify the new language addition  still.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":15,"context_line":"critical infrastructure, the need for parallelism in CPU-bound cryptographic"},{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"},{"line_number":17,"context_line":"for embedded distributed consensus, a recurring pattern of authorization-logic"},{"line_number":18,"context_line":"and concurrency-safety bugs in the current implementation, and the need to"},{"line_number":19,"context_line":"build modern, crypto-heavy authentication capabilities - a native OAuth2/OIDC"},{"line_number":20,"context_line":"provider and WebAuthn/passkey support - that Keystone lacks today and must"},{"line_number":21,"context_line":"build regardless of language."}],"source_content_type":"text/x-rst","patch_set":3,"id":"9f1118f3_30e69fc2","line":18,"range":{"start_line":18,"start_character":4,"end_line":18,"end_character":57},"in_reply_to":"a59aafc9_ff42595a","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"},{"line_number":17,"context_line":"for embedded distributed consensus, a recurring pattern of authorization-logic"},{"line_number":18,"context_line":"and concurrency-safety bugs in the current implementation, and the need to"},{"line_number":19,"context_line":"build modern, crypto-heavy authentication capabilities - a native OAuth2/OIDC"},{"line_number":20,"context_line":"provider and WebAuthn/passkey support - that Keystone lacks today and must"},{"line_number":21,"context_line":"build regardless of language."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"This proposal is sponsored by the Keystone project team, which commits to"}],"source_content_type":"text/x-rst","patch_set":3,"id":"2e4f4019_ef632df9","line":20,"range":{"start_line":19,"start_character":57,"end_line":20,"end_character":37},"updated":"2026-10-20 17:50:14.000000000","message":"How python is a blocker here?","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":16,"context_line":"workloads, the goal of predictable resource utilization at scale, the demand"},{"line_number":17,"context_line":"for embedded distributed consensus, a recurring pattern of authorization-logic"},{"line_number":18,"context_line":"and concurrency-safety bugs in the current implementation, and the need to"},{"line_number":19,"context_line":"build modern, crypto-heavy authentication capabilities - a native OAuth2/OIDC"},{"line_number":20,"context_line":"provider and WebAuthn/passkey support - that Keystone lacks today and must"},{"line_number":21,"context_line":"build regardless of language."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"This proposal is sponsored by the Keystone project team, which commits to"}],"source_content_type":"text/x-rst","patch_set":3,"id":"483b72e7_9a4990fa","line":20,"range":{"start_line":19,"start_character":57,"end_line":20,"end_character":37},"in_reply_to":"2e4f4019_ef632df9","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":20,"context_line":"provider and WebAuthn/passkey support - that Keystone lacks today and must"},{"line_number":21,"context_line":"build regardless of language."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"This proposal is sponsored by the Keystone project team, which commits to"},{"line_number":24,"context_line":"leading the Step 2 work should the TC approve this use case, including CI"},{"line_number":25,"context_line":"pipeline setup, packaging, and release integration for any Rust components it"},{"line_number":26,"context_line":"introduces."}],"source_content_type":"text/x-rst","patch_set":3,"id":"4c6deeea_76c97790","line":23,"range":{"start_line":23,"start_character":0,"end_line":23,"end_character":55},"updated":"2026-10-20 17:50:14.000000000","message":"It\u0027s worth mentioning, that for me a strict requirement here must be that any project written in a different language must be a new project, and pass all requirements to the new project and project team.\n\nOtherwise such effort feels like emotional blackmail/ultimatum.\n\nSo if the new language added, new project and project teams might be on-boarded to it, while existing projects must stay under the current governance and teams.\nWith all demands to the new project applied.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":20,"context_line":"provider and WebAuthn/passkey support - that Keystone lacks today and must"},{"line_number":21,"context_line":"build regardless of language."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"This proposal is sponsored by the Keystone project team, which commits to"},{"line_number":24,"context_line":"leading the Step 2 work should the TC approve this use case, including CI"},{"line_number":25,"context_line":"pipeline setup, packaging, and release integration for any Rust components it"},{"line_number":26,"context_line":"introduces."}],"source_content_type":"text/x-rst","patch_set":3,"id":"8975d272_902b4a48","line":23,"range":{"start_line":23,"start_character":0,"end_line":23,"end_character":55},"in_reply_to":"4c6deeea_76c97790","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":26,"context_line":"introduces."},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"The need for this addition was discussed on the mailing list starting May 11,"},{"line_number":29,"context_line":"2026 [1]_, drawing many replies. The thread tapered off without a formal close;"},{"line_number":30,"context_line":"this resolution is brought to the TC to reach the decision the mailing list"},{"line_number":31,"context_line":"discussion did not conclude on its own, and any objections raised there should"},{"line_number":32,"context_line":"be revisited during TC review."}],"source_content_type":"text/x-rst","patch_set":3,"id":"63b3297e_dbbc9703","line":29,"range":{"start_line":29,"start_character":66,"end_line":29,"end_character":78},"updated":"2026-10-30 06:32:06.000000000","message":"thanks for noting this; we could note there was a pretty lengthy debate (please note when it tapered off, or how many responses you gathered to qualify this if possible) and several supporting and opposing voices. I didn\u0027t expect this sort of an email debate to conclude decisively, but, it gave the community (code contributors, maintainers, packagers, operators) a way to voice their feelings.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":26,"context_line":"introduces."},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"The need for this addition was discussed on the mailing list starting May 11,"},{"line_number":29,"context_line":"2026 [1]_, drawing many replies. The thread tapered off without a formal close;"},{"line_number":30,"context_line":"this resolution is brought to the TC to reach the decision the mailing list"},{"line_number":31,"context_line":"discussion did not conclude on its own, and any objections raised there should"},{"line_number":32,"context_line":"be revisited during TC review."}],"source_content_type":"text/x-rst","patch_set":3,"id":"6dffc64b_f58db236","line":29,"range":{"start_line":29,"start_character":66,"end_line":29,"end_character":78},"in_reply_to":"63b3297e_dbbc9703","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":37,"context_line":"As the entry point and foundational authority for every API transaction across"},{"line_number":38,"context_line":"an OpenStack cloud, the identity service (Keystone) operates under unique,"},{"line_number":39,"context_line":"high-stress constraints. Under massive enterprise and public cloud deployments,"},{"line_number":40,"context_line":"the current Python-based architectural pattern—relying on Eventlet-driven"},{"line_number":41,"context_line":"cooperative multitasking under traditional synchronous WSGI server"},{"line_number":42,"context_line":"deployments—encounters systemic boundaries that cannot be cleanly resolved"},{"line_number":43,"context_line":"within the limits of the runtime:"},{"line_number":44,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3b4c6923_1bac699b","line":41,"range":{"start_line":40,"start_character":58,"end_line":41,"end_character":30},"updated":"2026-10-20 17:50:14.000000000","message":"I believe we have dropped Eventlet dependency across almost all services as of today.\nWhy is Keystone uniquely depend on Eventlet in a way no other OpenStack service does?","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":37,"context_line":"As the entry point and foundational authority for every API transaction across"},{"line_number":38,"context_line":"an OpenStack cloud, the identity service (Keystone) operates under unique,"},{"line_number":39,"context_line":"high-stress constraints. Under massive enterprise and public cloud deployments,"},{"line_number":40,"context_line":"the current Python-based architectural pattern—relying on Eventlet-driven"},{"line_number":41,"context_line":"cooperative multitasking under traditional synchronous WSGI server"},{"line_number":42,"context_line":"deployments—encounters systemic boundaries that cannot be cleanly resolved"},{"line_number":43,"context_line":"within the limits of the runtime:"},{"line_number":44,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"1f6e9752_cc6111c9","line":41,"range":{"start_line":40,"start_character":58,"end_line":41,"end_character":30},"in_reply_to":"2a54d5cf_66280f26","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":37,"context_line":"As the entry point and foundational authority for every API transaction across"},{"line_number":38,"context_line":"an OpenStack cloud, the identity service (Keystone) operates under unique,"},{"line_number":39,"context_line":"high-stress constraints. Under massive enterprise and public cloud deployments,"},{"line_number":40,"context_line":"the current Python-based architectural pattern—relying on Eventlet-driven"},{"line_number":41,"context_line":"cooperative multitasking under traditional synchronous WSGI server"},{"line_number":42,"context_line":"deployments—encounters systemic boundaries that cannot be cleanly resolved"},{"line_number":43,"context_line":"within the limits of the runtime:"},{"line_number":44,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"2a54d5cf_66280f26","line":41,"range":{"start_line":40,"start_character":58,"end_line":41,"end_character":30},"in_reply_to":"3b4c6923_1bac699b","updated":"2026-10-30 06:32:06.000000000","message":"+1; and the goal to drop eventlet from _everywhere_ in OpenStack will wrap up in 2027.2. Keystone eliminated eventlet from its requirements a while ago? So, this isn\u0027t a problem anymore.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":45,"context_line":"Compile-Time Memory Safety and Regulatory Compliance"},{"line_number":46,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"Global cybersecurity agencies (such as US CISA guidance [3]_) are increasingly"},{"line_number":49,"context_line":"calling out memory-unsafe languages as a systemic risk in critical"},{"line_number":50,"context_line":"infrastructure, and expect vendors to demonstrate a credible migration path."},{"line_number":51,"context_line":"While Python application code is intrinsically memory-safe due to its managed"},{"line_number":52,"context_line":"runtime, its underlying execution environment relies heavily on low-level C"},{"line_number":53,"context_line":"extensions for performance-critical segments - specifically in cryptography,"}],"source_content_type":"text/x-rst","patch_set":3,"id":"802f2ca7_9cdea648","line":50,"range":{"start_line":48,"start_character":0,"end_line":50,"end_character":14},"updated":"2026-10-20 17:50:14.000000000","message":"As you said - Python is not memory unsafe, so unsure how this sentence is relevant here.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":45,"context_line":"Compile-Time Memory Safety and Regulatory Compliance"},{"line_number":46,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"Global cybersecurity agencies (such as US CISA guidance [3]_) are increasingly"},{"line_number":49,"context_line":"calling out memory-unsafe languages as a systemic risk in critical"},{"line_number":50,"context_line":"infrastructure, and expect vendors to demonstrate a credible migration path."},{"line_number":51,"context_line":"While Python application code is intrinsically memory-safe due to its managed"},{"line_number":52,"context_line":"runtime, its underlying execution environment relies heavily on low-level C"},{"line_number":53,"context_line":"extensions for performance-critical segments - specifically in cryptography,"}],"source_content_type":"text/x-rst","patch_set":3,"id":"9c2b91d8_91592b49","line":50,"range":{"start_line":48,"start_character":0,"end_line":50,"end_character":14},"in_reply_to":"802f2ca7_9cdea648","updated":"2026-10-30 06:32:06.000000000","message":"It isn\u0027t required, the explanation below doesn\u0027t sound helpful either. \n\nLink at [3] is now gone/404. \n\nBut, the CISA/CRA recommendations are targeting C/C++.. they explicitly say this:\n\n\"\"\"\nThese examples underscore the\nurgency of finding better solutions. MSLs such as Ada, C#, Delphi/Object Pascal, Go,\nJava, Python, Ruby, Rust, and Swift offer built-in protections against memory safety\nissues, making them a strategic choice for developing more secure software.\n\"\"\"\n\nhttps://media.defense.gov/2025/Jun/23/2003742198/-1/-1/0/CSI_MEMORY_SAFE_LANGUAGES_REDUCING_VULNERABILITIES_IN_MODERN_SOFTWARE_DEVELOPMENT.PDF","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":45,"context_line":"Compile-Time Memory Safety and Regulatory Compliance"},{"line_number":46,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"Global cybersecurity agencies (such as US CISA guidance [3]_) are increasingly"},{"line_number":49,"context_line":"calling out memory-unsafe languages as a systemic risk in critical"},{"line_number":50,"context_line":"infrastructure, and expect vendors to demonstrate a credible migration path."},{"line_number":51,"context_line":"While Python application code is intrinsically memory-safe due to its managed"},{"line_number":52,"context_line":"runtime, its underlying execution environment relies heavily on low-level C"},{"line_number":53,"context_line":"extensions for performance-critical segments - specifically in cryptography,"}],"source_content_type":"text/x-rst","patch_set":3,"id":"10f4c4bf_5d809972","line":50,"range":{"start_line":48,"start_character":0,"end_line":50,"end_character":14},"in_reply_to":"9c2b91d8_91592b49","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":49,"context_line":"calling out memory-unsafe languages as a systemic risk in critical"},{"line_number":50,"context_line":"infrastructure, and expect vendors to demonstrate a credible migration path."},{"line_number":51,"context_line":"While Python application code is intrinsically memory-safe due to its managed"},{"line_number":52,"context_line":"runtime, its underlying execution environment relies heavily on low-level C"},{"line_number":53,"context_line":"extensions for performance-critical segments - specifically in cryptography,"},{"line_number":54,"context_line":"token serialization, and database driver bindings. This creates a fragmented"},{"line_number":55,"context_line":"audit boundary. Verifying the integrity of these native dependency trees under"},{"line_number":56,"context_line":"standard Python tooling is a persistent operational challenge. Rust\u0027s"}],"source_content_type":"text/x-rst","patch_set":3,"id":"949a6985_a58d8dc4","line":53,"range":{"start_line":52,"start_character":13,"end_line":53,"end_character":75},"updated":"2026-10-20 17:50:14.000000000","message":"I think python-cryptography is rust-based, not C based. So this is not even true today.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"2a49234304a9a4a28142c28e1d935d9bb49afa95","unresolved":true,"context_lines":[{"line_number":48,"context_line":"Global cybersecurity agencies (such as US CISA guidance [3]_) are increasingly"},{"line_number":49,"context_line":"calling out memory-unsafe languages as a systemic risk in critical"},{"line_number":50,"context_line":"infrastructure, and expect vendors to demonstrate a credible migration path."},{"line_number":51,"context_line":"While Python application code is intrinsically memory-safe due to its managed"},{"line_number":52,"context_line":"runtime, its underlying execution environment relies heavily on low-level C"},{"line_number":53,"context_line":"extensions for performance-critical segments - specifically in cryptography,"},{"line_number":54,"context_line":"token serialization, and database driver bindings. This creates a fragmented"},{"line_number":55,"context_line":"audit boundary. Verifying the integrity of these native dependency trees under"},{"line_number":56,"context_line":"standard Python tooling is a persistent operational challenge. Rust\u0027s"}],"source_content_type":"text/x-rst","patch_set":3,"id":"c20ae0b8_d8b24afe","line":53,"range":{"start_line":51,"start_character":0,"end_line":53,"end_character":44},"updated":"2026-10-20 18:02:05.000000000","message":"To me this is like saying \"while Rust is intrinsically memory-safe, the x86 assembly that it ultimately relies on is not\". This may be a problem for Python itself, or the various crypto libraries, but from a regulatory perspective I can\u0027t imagine distributors of openstack code have to care about the entire stack down to the microcode if they can point to someone else responsible (which may be themselves, granted, but likely handled by different teams).","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":49,"context_line":"calling out memory-unsafe languages as a systemic risk in critical"},{"line_number":50,"context_line":"infrastructure, and expect vendors to demonstrate a credible migration path."},{"line_number":51,"context_line":"While Python application code is intrinsically memory-safe due to its managed"},{"line_number":52,"context_line":"runtime, its underlying execution environment relies heavily on low-level C"},{"line_number":53,"context_line":"extensions for performance-critical segments - specifically in cryptography,"},{"line_number":54,"context_line":"token serialization, and database driver bindings. This creates a fragmented"},{"line_number":55,"context_line":"audit boundary. Verifying the integrity of these native dependency trees under"},{"line_number":56,"context_line":"standard Python tooling is a persistent operational challenge. Rust\u0027s"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3ba128b7_20022c0d","line":53,"range":{"start_line":52,"start_character":13,"end_line":53,"end_character":75},"in_reply_to":"949a6985_a58d8dc4","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":48,"context_line":"Global cybersecurity agencies (such as US CISA guidance [3]_) are increasingly"},{"line_number":49,"context_line":"calling out memory-unsafe languages as a systemic risk in critical"},{"line_number":50,"context_line":"infrastructure, and expect vendors to demonstrate a credible migration path."},{"line_number":51,"context_line":"While Python application code is intrinsically memory-safe due to its managed"},{"line_number":52,"context_line":"runtime, its underlying execution environment relies heavily on low-level C"},{"line_number":53,"context_line":"extensions for performance-critical segments - specifically in cryptography,"},{"line_number":54,"context_line":"token serialization, and database driver bindings. This creates a fragmented"},{"line_number":55,"context_line":"audit boundary. Verifying the integrity of these native dependency trees under"},{"line_number":56,"context_line":"standard Python tooling is a persistent operational challenge. Rust\u0027s"}],"source_content_type":"text/x-rst","patch_set":3,"id":"70c082a2_476ebc01","line":53,"range":{"start_line":51,"start_character":0,"end_line":53,"end_character":44},"in_reply_to":"c20ae0b8_d8b24afe","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":57,"context_line":"ownership model eliminates this class of vulnerability at compile time, without"},{"line_number":58,"context_line":"the runtime cost or non-deterministic pause times of a garbage-collected"},{"line_number":59,"context_line":"language such as Go - relevant given Keystone\u0027s latency-sensitive position on"},{"line_number":60,"context_line":"every API call."},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"Supply-chain transparency is a related regulatory driver. The EU Cyber"},{"line_number":63,"context_line":"Resilience Act [13]_ places software bill of materials (SBOM), third-party"}],"source_content_type":"text/x-rst","patch_set":3,"id":"168ab498_ac48c5b2","line":60,"range":{"start_line":60,"start_character":14,"end_line":60,"end_character":15},"updated":"2026-10-30 06:32:06.000000000","message":"This argument would be stronger if a regulator had actually flagged Keystone for being Python-based. Without a concrete compliance requirement from a real deployment, this is speculative.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":57,"context_line":"ownership model eliminates this class of vulnerability at compile time, without"},{"line_number":58,"context_line":"the runtime cost or non-deterministic pause times of a garbage-collected"},{"line_number":59,"context_line":"language such as Go - relevant given Keystone\u0027s latency-sensitive position on"},{"line_number":60,"context_line":"every API call."},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"Supply-chain transparency is a related regulatory driver. The EU Cyber"},{"line_number":63,"context_line":"Resilience Act [13]_ places software bill of materials (SBOM), third-party"}],"source_content_type":"text/x-rst","patch_set":3,"id":"c59b1caa_6b4fb0ce","line":60,"range":{"start_line":60,"start_character":14,"end_line":60,"end_character":15},"in_reply_to":"168ab498_ac48c5b2","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"2a49234304a9a4a28142c28e1d935d9bb49afa95","unresolved":true,"context_lines":[{"line_number":79,"context_line":"the RustSec advisory database. This comes with a cost: because dependencies"},{"line_number":80,"context_line":"are compiled in, fixing a vulnerable one requires rebuilding and redeploying"},{"line_number":81,"context_line":"every affected binary, so Step 2 must include a rebuild and release process"},{"line_number":82,"context_line":"able to meet the Act\u0027s security-update timelines."},{"line_number":83,"context_line":""},{"line_number":84,"context_line":"Parallelism and Cryptographic Throughput under the GIL"},{"line_number":85,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":3,"id":"8804d81e_cb438145","line":82,"updated":"2026-10-20 18:02:05.000000000","message":"Aside from the fact that all these CRA requirements don\u0027t apply to the upstream community, I feel like this is a problematic justification for (at least) two reasons:\n\n1. Seems to me you\u0027re making an argument for re-writing all of openstack in another language. In the absence of that, I don\u0027t see how this argument is meaningful for an otherwise-entirely-python ecosystem of related components that most people see as a single consumable unit.\n\n2. I feel like the SBOM is entirely clear when packaged in something like a container or a readonly-root sort of environment. Both of these things are packager, distributor, vendor issues and we need not make high-level decisions for the entire community based on the dynamic-ness of the (otherwise very popular) platform we\u0027re using.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":79,"context_line":"the RustSec advisory database. This comes with a cost: because dependencies"},{"line_number":80,"context_line":"are compiled in, fixing a vulnerable one requires rebuilding and redeploying"},{"line_number":81,"context_line":"every affected binary, so Step 2 must include a rebuild and release process"},{"line_number":82,"context_line":"able to meet the Act\u0027s security-update timelines."},{"line_number":83,"context_line":""},{"line_number":84,"context_line":"Parallelism and Cryptographic Throughput under the GIL"},{"line_number":85,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":3,"id":"b7c7f360_f9415c63","line":82,"in_reply_to":"86798d69_29ef5543","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"b9cf2365cb8234a5354893d26b5bb4e3848d86ec","unresolved":true,"context_lines":[{"line_number":79,"context_line":"the RustSec advisory database. This comes with a cost: because dependencies"},{"line_number":80,"context_line":"are compiled in, fixing a vulnerable one requires rebuilding and redeploying"},{"line_number":81,"context_line":"every affected binary, so Step 2 must include a rebuild and release process"},{"line_number":82,"context_line":"able to meet the Act\u0027s security-update timelines."},{"line_number":83,"context_line":""},{"line_number":84,"context_line":"Parallelism and Cryptographic Throughput under the GIL"},{"line_number":85,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":3,"id":"a2519d6e_3ae0766d","line":82,"in_reply_to":"8804d81e_cb438145","updated":"2026-10-22 17:02:24.000000000","message":"I disagree strongly with a position, that upstream should not care about that. For something like SDK it does not matter. Keystone, or IAM, is the first thing the auditor is looking at, so when we as upstream make it easier for operator everybody wins. \nI personally was involved in few tenders where this was so much of a point.\nYou are right, SBOM for container is easy. That is what resolution states, but OpenStack can be run not only inside container, I wanted to explain why it differs.\nAs stated few lines above, for Rust app you can embed SBOM into the binary and let the process inspect itself and refuse to start when there is a known vulnerability. For IAM component this is what auditor is happy to hear rather than - we have sbom for some image, but we can\u0027t prove you this is the image that runs now in prod right now. Some customers and auditors, at least here in EU, are requesting very different and specific things. As upstream we must care and make it easier for operators, otherwise we loose them.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":79,"context_line":"the RustSec advisory database. This comes with a cost: because dependencies"},{"line_number":80,"context_line":"are compiled in, fixing a vulnerable one requires rebuilding and redeploying"},{"line_number":81,"context_line":"every affected binary, so Step 2 must include a rebuild and release process"},{"line_number":82,"context_line":"able to meet the Act\u0027s security-update timelines."},{"line_number":83,"context_line":""},{"line_number":84,"context_line":"Parallelism and Cryptographic Throughput under the GIL"},{"line_number":85,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":3,"id":"86798d69_29ef5543","line":82,"in_reply_to":"a2519d6e_3ae0766d","updated":"2026-10-30 06:32:06.000000000","message":"devils\u0027 advocate -   doesn\u0027t embedding the SBOM in the binary limit downstream distributors\u0027 ability to customize or patch?\n\nThe deeper issue is that Dan\u0027s right - we can\u0027t set the precedent that OpenStack cannot meet CRA compliance because it is written in python today. Your argument that keystone is different can be made about other security sensitive portions of OpenStack as well.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":87,"context_line":"In a high-density OpenStack environment, a single user-facing action often"},{"line_number":88,"context_line":"triggers a cascading sequence of internal service-to-service API calls, each"},{"line_number":89,"context_line":"requiring independent token validation, cryptographic signature verification,"},{"line_number":90,"context_line":"and granular policy engine checks. While Python\u0027s greenlet-based cooperative"},{"line_number":91,"context_line":"threading handles standard network socket I/O effectively, it falls short when"},{"line_number":92,"context_line":"executing CPU-bound cryptographic operations. Because compute-heavy validation"},{"line_number":93,"context_line":"paths cannot be asynchronously yielded by the Eventlet hub, a single complex"},{"line_number":94,"context_line":"token verification blocks the execution context. To counter this, operators"},{"line_number":95,"context_line":"are forced to deploy heavily scaled multi-process WSGI clusters - a"},{"line_number":96,"context_line":"configuration that trades memory headroom for concurrency. At sufficient"},{"line_number":97,"context_line":"scale, this conflicts directly with the memory the OS needs for filesystem"}],"source_content_type":"text/x-rst","patch_set":3,"id":"66475791_17dced2c","line":94,"range":{"start_line":90,"start_character":35,"end_line":94,"end_character":48},"updated":"2026-10-20 17:50:14.000000000","message":"I think this statement is outdated and does not address changes in Python\u003e\u003d3.13\nI don\u0027t think we should be looking at the language past while determening the future.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":87,"context_line":"In a high-density OpenStack environment, a single user-facing action often"},{"line_number":88,"context_line":"triggers a cascading sequence of internal service-to-service API calls, each"},{"line_number":89,"context_line":"requiring independent token validation, cryptographic signature verification,"},{"line_number":90,"context_line":"and granular policy engine checks. While Python\u0027s greenlet-based cooperative"},{"line_number":91,"context_line":"threading handles standard network socket I/O effectively, it falls short when"},{"line_number":92,"context_line":"executing CPU-bound cryptographic operations. Because compute-heavy validation"},{"line_number":93,"context_line":"paths cannot be asynchronously yielded by the Eventlet hub, a single complex"},{"line_number":94,"context_line":"token verification blocks the execution context. To counter this, operators"},{"line_number":95,"context_line":"are forced to deploy heavily scaled multi-process WSGI clusters - a"},{"line_number":96,"context_line":"configuration that trades memory headroom for concurrency. At sufficient"},{"line_number":97,"context_line":"scale, this conflicts directly with the memory the OS needs for filesystem"}],"source_content_type":"text/x-rst","patch_set":3,"id":"d54e41e1_13118171","line":94,"range":{"start_line":90,"start_character":35,"end_line":94,"end_character":48},"in_reply_to":"66475791_17dced2c","updated":"2026-10-30 06:32:06.000000000","message":"plus keystone doesn\u0027t support eventlet hub anymore?","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":87,"context_line":"In a high-density OpenStack environment, a single user-facing action often"},{"line_number":88,"context_line":"triggers a cascading sequence of internal service-to-service API calls, each"},{"line_number":89,"context_line":"requiring independent token validation, cryptographic signature verification,"},{"line_number":90,"context_line":"and granular policy engine checks. While Python\u0027s greenlet-based cooperative"},{"line_number":91,"context_line":"threading handles standard network socket I/O effectively, it falls short when"},{"line_number":92,"context_line":"executing CPU-bound cryptographic operations. Because compute-heavy validation"},{"line_number":93,"context_line":"paths cannot be asynchronously yielded by the Eventlet hub, a single complex"},{"line_number":94,"context_line":"token verification blocks the execution context. To counter this, operators"},{"line_number":95,"context_line":"are forced to deploy heavily scaled multi-process WSGI clusters - a"},{"line_number":96,"context_line":"configuration that trades memory headroom for concurrency. At sufficient"},{"line_number":97,"context_line":"scale, this conflicts directly with the memory the OS needs for filesystem"}],"source_content_type":"text/x-rst","patch_set":3,"id":"ab522c39_2f485505","line":94,"range":{"start_line":90,"start_character":35,"end_line":94,"end_character":48},"in_reply_to":"d54e41e1_13118171","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":98,"context_line":"buffer caches and database page pools, adding pressure on overall"},{"line_number":99,"context_line":"control-plane throughput."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"The mismatch between Eventlet\u0027s greenlet model and standard concurrency"},{"line_number":102,"context_line":"primitives is not only a throughput concern. Bug 1360446 [11]_ found that"},{"line_number":103,"context_line":"python-memcached\u0027s use of threadlocal storage for connection state, when"},{"line_number":104,"context_line":"combined with Eventlet\u0027s greenlet-based concurrency, prevented memcached"}],"source_content_type":"text/x-rst","patch_set":3,"id":"16f94cf8_dd8f2f6a","line":101,"range":{"start_line":101,"start_character":21,"end_line":101,"end_character":46},"updated":"2026-10-20 17:50:14.000000000","message":"Eventlet is dropped from majority of services today. Is that still relevant?","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":98,"context_line":"buffer caches and database page pools, adding pressure on overall"},{"line_number":99,"context_line":"control-plane throughput."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"The mismatch between Eventlet\u0027s greenlet model and standard concurrency"},{"line_number":102,"context_line":"primitives is not only a throughput concern. Bug 1360446 [11]_ found that"},{"line_number":103,"context_line":"python-memcached\u0027s use of threadlocal storage for connection state, when"},{"line_number":104,"context_line":"combined with Eventlet\u0027s greenlet-based concurrency, prevented memcached"}],"source_content_type":"text/x-rst","patch_set":3,"id":"a7c7ce1d_c0ccfbe1","line":101,"range":{"start_line":101,"start_character":21,"end_line":101,"end_character":46},"in_reply_to":"16f94cf8_dd8f2f6a","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"2a49234304a9a4a28142c28e1d935d9bb49afa95","unresolved":true,"context_lines":[{"line_number":109,"context_line":"built into a common Python library. Rust\u0027s async runtimes are built around a"},{"line_number":110,"context_line":"single, explicit concurrency model rather than layering cooperative"},{"line_number":111,"context_line":"greenlets on top of libraries written against OS threads, removing this class"},{"line_number":112,"context_line":"of mismatch."},{"line_number":113,"context_line":""},{"line_number":114,"context_line":"This is compounded by the state of Keystone\u0027s own web framework. Keystone\u0027s"},{"line_number":115,"context_line":"REST layer is built on Flask-RESTful, which has had no release in roughly two"}],"source_content_type":"text/x-rst","patch_set":3,"id":"e044783e_1132fc27","line":112,"updated":"2026-10-20 18:02:05.000000000","message":"Most everyone is off or close to fully transitioning off of eventlet at this point. Staying on it is not an option, nor is it even in keystone\u0027s current requirements.txt. Is this a general statement or actually related to the question of keystone specifically?","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":109,"context_line":"built into a common Python library. Rust\u0027s async runtimes are built around a"},{"line_number":110,"context_line":"single, explicit concurrency model rather than layering cooperative"},{"line_number":111,"context_line":"greenlets on top of libraries written against OS threads, removing this class"},{"line_number":112,"context_line":"of mismatch."},{"line_number":113,"context_line":""},{"line_number":114,"context_line":"This is compounded by the state of Keystone\u0027s own web framework. Keystone\u0027s"},{"line_number":115,"context_line":"REST layer is built on Flask-RESTful, which has had no release in roughly two"}],"source_content_type":"text/x-rst","patch_set":3,"id":"817b93c5_7f99e089","line":112,"in_reply_to":"1010e79b_6fd561c9","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":109,"context_line":"built into a common Python library. Rust\u0027s async runtimes are built around a"},{"line_number":110,"context_line":"single, explicit concurrency model rather than layering cooperative"},{"line_number":111,"context_line":"greenlets on top of libraries written against OS threads, removing this class"},{"line_number":112,"context_line":"of mismatch."},{"line_number":113,"context_line":""},{"line_number":114,"context_line":"This is compounded by the state of Keystone\u0027s own web framework. Keystone\u0027s"},{"line_number":115,"context_line":"REST layer is built on Flask-RESTful, which has had no release in roughly two"}],"source_content_type":"text/x-rst","patch_set":3,"id":"1010e79b_6fd561c9","line":112,"in_reply_to":"e044783e_1132fc27","updated":"2026-10-30 06:32:06.000000000","message":"+1 on dropping this historic artifact; eventlet is gone from keystone.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":112,"context_line":"of mismatch."},{"line_number":113,"context_line":""},{"line_number":114,"context_line":"This is compounded by the state of Keystone\u0027s own web framework. Keystone\u0027s"},{"line_number":115,"context_line":"REST layer is built on Flask-RESTful, which has had no release in roughly two"},{"line_number":116,"context_line":"years and has an open, long-standing \"flaskification\" migration bug"},{"line_number":117,"context_line":"(bug 1776504 [12]_) against it; it was never designed with native async"},{"line_number":118,"context_line":"support, consistent with Keystone\u0027s current synchronous, Eventlet-driven"}],"source_content_type":"text/x-rst","patch_set":3,"id":"cae5839b_b705e5d0","line":115,"range":{"start_line":115,"start_character":23,"end_line":115,"end_character":36},"updated":"2026-10-20 17:50:14.000000000","message":"with that last flask version was released in Feb 2026. While indeed flask-restful looks dead, I\u0027d assume is relatively small set of extra utils to replace if needed.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":112,"context_line":"of mismatch."},{"line_number":113,"context_line":""},{"line_number":114,"context_line":"This is compounded by the state of Keystone\u0027s own web framework. Keystone\u0027s"},{"line_number":115,"context_line":"REST layer is built on Flask-RESTful, which has had no release in roughly two"},{"line_number":116,"context_line":"years and has an open, long-standing \"flaskification\" migration bug"},{"line_number":117,"context_line":"(bug 1776504 [12]_) against it; it was never designed with native async"},{"line_number":118,"context_line":"support, consistent with Keystone\u0027s current synchronous, Eventlet-driven"}],"source_content_type":"text/x-rst","patch_set":3,"id":"1fa63254_3e36c9b4","line":115,"range":{"start_line":115,"start_character":23,"end_line":115,"end_character":36},"in_reply_to":"cae5839b_b705e5d0","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":115,"context_line":"REST layer is built on Flask-RESTful, which has had no release in roughly two"},{"line_number":116,"context_line":"years and has an open, long-standing \"flaskification\" migration bug"},{"line_number":117,"context_line":"(bug 1776504 [12]_) against it; it was never designed with native async"},{"line_number":118,"context_line":"support, consistent with Keystone\u0027s current synchronous, Eventlet-driven"},{"line_number":119,"context_line":"request handling. A move to a modern, async-native stack is therefore not"},{"line_number":120,"context_line":"purely a language choice added on top of working tooling - part of the"},{"line_number":121,"context_line":"underlying Python web framework is already due for replacement regardless."},{"line_number":122,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"ac02a14f_553a1055","line":119,"range":{"start_line":118,"start_character":57,"end_line":119,"end_character":16},"updated":"2026-10-30 06:32:06.000000000","message":"not true anymore, i.e., you don\u0027t have this as a choice","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":115,"context_line":"REST layer is built on Flask-RESTful, which has had no release in roughly two"},{"line_number":116,"context_line":"years and has an open, long-standing \"flaskification\" migration bug"},{"line_number":117,"context_line":"(bug 1776504 [12]_) against it; it was never designed with native async"},{"line_number":118,"context_line":"support, consistent with Keystone\u0027s current synchronous, Eventlet-driven"},{"line_number":119,"context_line":"request handling. A move to a modern, async-native stack is therefore not"},{"line_number":120,"context_line":"purely a language choice added on top of working tooling - part of the"},{"line_number":121,"context_line":"underlying Python web framework is already due for replacement regardless."},{"line_number":122,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"a504ed18_9cf30518","line":119,"range":{"start_line":118,"start_character":57,"end_line":119,"end_character":16},"in_reply_to":"ac02a14f_553a1055","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"2a49234304a9a4a28142c28e1d935d9bb49afa95","unresolved":true,"context_lines":[{"line_number":118,"context_line":"support, consistent with Keystone\u0027s current synchronous, Eventlet-driven"},{"line_number":119,"context_line":"request handling. A move to a modern, async-native stack is therefore not"},{"line_number":120,"context_line":"purely a language choice added on top of working tooling - part of the"},{"line_number":121,"context_line":"underlying Python web framework is already due for replacement regardless."},{"line_number":122,"context_line":""},{"line_number":123,"context_line":"Distributed Consensus and Local State Resilience at Scale"},{"line_number":124,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":3,"id":"d57fcbfa_4ba72192","line":121,"updated":"2026-10-20 18:02:05.000000000","message":"I think a number of other projects are probably in this boat as well, as will anything in a few years that picks any framework today.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":118,"context_line":"support, consistent with Keystone\u0027s current synchronous, Eventlet-driven"},{"line_number":119,"context_line":"request handling. A move to a modern, async-native stack is therefore not"},{"line_number":120,"context_line":"purely a language choice added on top of working tooling - part of the"},{"line_number":121,"context_line":"underlying Python web framework is already due for replacement regardless."},{"line_number":122,"context_line":""},{"line_number":123,"context_line":"Distributed Consensus and Local State Resilience at Scale"},{"line_number":124,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":3,"id":"00919e2b_1eaa6f8a","line":121,"in_reply_to":"870e2cf3_eecc0c32","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":118,"context_line":"support, consistent with Keystone\u0027s current synchronous, Eventlet-driven"},{"line_number":119,"context_line":"request handling. A move to a modern, async-native stack is therefore not"},{"line_number":120,"context_line":"purely a language choice added on top of working tooling - part of the"},{"line_number":121,"context_line":"underlying Python web framework is already due for replacement regardless."},{"line_number":122,"context_line":""},{"line_number":123,"context_line":"Distributed Consensus and Local State Resilience at Scale"},{"line_number":124,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":3,"id":"870e2cf3_eecc0c32","line":121,"in_reply_to":"d57fcbfa_4ba72192","updated":"2026-10-30 06:32:06.000000000","message":"/me agrees on betting on framework longevity :/","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":139,"context_line":"Authorization Logic Correctness"},{"line_number":140,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":141,"context_line":""},{"line_number":142,"context_line":"Keystone\u0027s recent security advisories illustrate a distinct class of problem"},{"line_number":143,"context_line":"that memory safety does not address: gaps in credential and trust state"},{"line_number":144,"context_line":"handling. OSSA-2025-002 [4]_ allowed unauthenticated EC2/S3 token requests to"},{"line_number":145,"context_line":"grant Keystone authorization; OSSA-2026-005 [5]_ let restricted application"},{"line_number":146,"context_line":"credentials create EC2 credentials outside their intended scope; and"}],"source_content_type":"text/x-rst","patch_set":3,"id":"b0867433_f9d422e9","line":143,"range":{"start_line":142,"start_character":0,"end_line":143,"end_character":35},"updated":"2026-10-20 17:50:14.000000000","message":"neither would a different language.\nsecurity vulnarabilities are commonly found in rust re-writes.\nNice read show-casing that this argument can hardly be used for a different language.\n\nhttps://www.decryptiondigest.com/blog/does-rewriting-rust-prevent-ai-discovered-vulnerabilities","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":139,"context_line":"Authorization Logic Correctness"},{"line_number":140,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":141,"context_line":""},{"line_number":142,"context_line":"Keystone\u0027s recent security advisories illustrate a distinct class of problem"},{"line_number":143,"context_line":"that memory safety does not address: gaps in credential and trust state"},{"line_number":144,"context_line":"handling. OSSA-2025-002 [4]_ allowed unauthenticated EC2/S3 token requests to"},{"line_number":145,"context_line":"grant Keystone authorization; OSSA-2026-005 [5]_ let restricted application"},{"line_number":146,"context_line":"credentials create EC2 credentials outside their intended scope; and"}],"source_content_type":"text/x-rst","patch_set":3,"id":"f70367d4_21271a45","line":143,"range":{"start_line":142,"start_character":0,"end_line":143,"end_character":35},"in_reply_to":"b0867433_f9d422e9","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":140,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":141,"context_line":""},{"line_number":142,"context_line":"Keystone\u0027s recent security advisories illustrate a distinct class of problem"},{"line_number":143,"context_line":"that memory safety does not address: gaps in credential and trust state"},{"line_number":144,"context_line":"handling. OSSA-2025-002 [4]_ allowed unauthenticated EC2/S3 token requests to"},{"line_number":145,"context_line":"grant Keystone authorization; OSSA-2026-005 [5]_ let restricted application"},{"line_number":146,"context_line":"credentials create EC2 credentials outside their intended scope; and"},{"line_number":147,"context_line":"OSSA-2026-015 [6]_ covers multiple related issues, including application"}],"source_content_type":"text/x-rst","patch_set":3,"id":"f3ee8406_a5be424b","line":144,"range":{"start_line":143,"start_character":37,"end_line":144,"end_character":9},"updated":"2026-10-30 06:32:06.000000000","message":"With my VMT hat on, this statement hits home for me. I agree wholeheartedly on re-architecting some of these half-implemented/poorly propagated features. Nearly every vulnerability we triaged with you in this year has identified things like this. \n\nThe details that follow this are not really useful in the rust discussion though. \n\nFurther below you speak of nullable-field states. Strict type checking can now be implemented to close such gaps in python","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":140,"context_line":"~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"},{"line_number":141,"context_line":""},{"line_number":142,"context_line":"Keystone\u0027s recent security advisories illustrate a distinct class of problem"},{"line_number":143,"context_line":"that memory safety does not address: gaps in credential and trust state"},{"line_number":144,"context_line":"handling. OSSA-2025-002 [4]_ allowed unauthenticated EC2/S3 token requests to"},{"line_number":145,"context_line":"grant Keystone authorization; OSSA-2026-005 [5]_ let restricted application"},{"line_number":146,"context_line":"credentials create EC2 credentials outside their intended scope; and"},{"line_number":147,"context_line":"OSSA-2026-015 [6]_ covers multiple related issues, including application"}],"source_content_type":"text/x-rst","patch_set":3,"id":"d4a89d23_bdaaf8d4","line":144,"range":{"start_line":143,"start_character":37,"end_line":144,"end_character":9},"in_reply_to":"f3ee8406_a5be424b","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":158,"context_line":""},{"line_number":159,"context_line":"None of these are memory-safety bugs, and Python\u0027s managed runtime would not"},{"line_number":160,"context_line":"have prevented any of them - they are gaps in how credential, trust, token,"},{"line_number":161,"context_line":"and nullable-field state was modeled and validated, and in several cases the"},{"line_number":162,"context_line":"gap is literally an unhandled branch (a NULL case, or an ``else`` that was"},{"line_number":163,"context_line":"never written). Rust\u0027s sum types and exhaustive ``match`` expressions require"},{"line_number":164,"context_line":"every state and every combination of caller/owner/scope to be handled"}],"source_content_type":"text/x-rst","patch_set":3,"id":"f2b1fcc0_c91132d9","line":161,"range":{"start_line":161,"start_character":0,"end_line":161,"end_character":50},"updated":"2026-10-20 17:50:14.000000000","message":"the type checks can be enforced in the Python as of today and couple of projects have already adopted that.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":158,"context_line":""},{"line_number":159,"context_line":"None of these are memory-safety bugs, and Python\u0027s managed runtime would not"},{"line_number":160,"context_line":"have prevented any of them - they are gaps in how credential, trust, token,"},{"line_number":161,"context_line":"and nullable-field state was modeled and validated, and in several cases the"},{"line_number":162,"context_line":"gap is literally an unhandled branch (a NULL case, or an ``else`` that was"},{"line_number":163,"context_line":"never written). Rust\u0027s sum types and exhaustive ``match`` expressions require"},{"line_number":164,"context_line":"every state and every combination of caller/owner/scope to be handled"}],"source_content_type":"text/x-rst","patch_set":3,"id":"515858c1_80f6f6e6","line":161,"range":{"start_line":161,"start_character":0,"end_line":161,"end_character":50},"in_reply_to":"f2b1fcc0_c91132d9","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":184,"context_line":"- with a previous request\u0027s credentials still attached. The first proposed"},{"line_number":185,"context_line":"fix for that bug introduced its own regression, illustrating how hard it is"},{"line_number":186,"context_line":"to reason correctly about implicit, mutable, thread-local state by hand."},{"line_number":187,"context_line":"Rust\u0027s borrow checker rejects shared mutable state unless it is proven"},{"line_number":188,"context_line":"data-race-free at compile time, and its ownership model favors passing"},{"line_number":189,"context_line":"context explicitly by value or reference over implicit thread-local globals,"},{"line_number":190,"context_line":"making this class of bug structurally harder to introduce."}],"source_content_type":"text/x-rst","patch_set":3,"id":"9dccdc0e_2fab17cd","line":187,"range":{"start_line":187,"start_character":22,"end_line":187,"end_character":50},"updated":"2026-10-30 06:32:06.000000000","message":"a language safety against poor programmers? :)\n\nThe borrow checker sounds like it helps, but these are architectural issues that better tests and design reviews also catch.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":184,"context_line":"- with a previous request\u0027s credentials still attached. The first proposed"},{"line_number":185,"context_line":"fix for that bug introduced its own regression, illustrating how hard it is"},{"line_number":186,"context_line":"to reason correctly about implicit, mutable, thread-local state by hand."},{"line_number":187,"context_line":"Rust\u0027s borrow checker rejects shared mutable state unless it is proven"},{"line_number":188,"context_line":"data-race-free at compile time, and its ownership model favors passing"},{"line_number":189,"context_line":"context explicitly by value or reference over implicit thread-local globals,"},{"line_number":190,"context_line":"making this class of bug structurally harder to introduce."}],"source_content_type":"text/x-rst","patch_set":3,"id":"eca937ed_6d2ba6dd","line":187,"range":{"start_line":187,"start_character":22,"end_line":187,"end_character":50},"in_reply_to":"9dccdc0e_2fab17cd","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":208,"context_line":"parallelism section above. Since this work has to happen anyway, building it"},{"line_number":209,"context_line":"in Rust means the new, crypto-heavy OIDC implementation gets compile-time"},{"line_number":210,"context_line":"memory safety and better throughput from day one, instead of being built once"},{"line_number":211,"context_line":"in Python and rewritten later."},{"line_number":212,"context_line":""},{"line_number":213,"context_line":"WebAuthn (passkey) support is another modern authentication capability"},{"line_number":214,"context_line":"Keystone does not have today, and operators are increasingly asking for it."}],"source_content_type":"text/x-rst","patch_set":3,"id":"8211990c_c4e81979","line":211,"range":{"start_line":211,"start_character":14,"end_line":211,"end_character":30},"updated":"2026-10-30 06:32:06.000000000","message":"circular argument. \n\nwe should do this now instead because we\u0027ll do it anyway later?","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":208,"context_line":"parallelism section above. Since this work has to happen anyway, building it"},{"line_number":209,"context_line":"in Rust means the new, crypto-heavy OIDC implementation gets compile-time"},{"line_number":210,"context_line":"memory safety and better throughput from day one, instead of being built once"},{"line_number":211,"context_line":"in Python and rewritten later."},{"line_number":212,"context_line":""},{"line_number":213,"context_line":"WebAuthn (passkey) support is another modern authentication capability"},{"line_number":214,"context_line":"Keystone does not have today, and operators are increasingly asking for it."}],"source_content_type":"text/x-rst","patch_set":3,"id":"305cdec6_d45a42ac","line":211,"range":{"start_line":211,"start_character":14,"end_line":211,"end_character":30},"in_reply_to":"8211990c_c4e81979","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":221,"context_line":"Proposed solution"},{"line_number":222,"context_line":"-----------------"},{"line_number":223,"context_line":""},{"line_number":224,"context_line":"The integration of Rust into the OpenStack ecosystem addresses these"},{"line_number":225,"context_line":"structural limitations. Rust delivers near-native execution speed and"},{"line_number":226,"context_line":"deterministic resource control without garbage-collection pauses, while"},{"line_number":227,"context_line":"enforcing memory safety and data-race prevention at compile time through its"},{"line_number":228,"context_line":"ownership model."}],"source_content_type":"text/x-rst","patch_set":3,"id":"b4af82ac_9ef97a7d","line":225,"range":{"start_line":224,"start_character":0,"end_line":225,"end_character":23},"updated":"2026-10-20 17:50:14.000000000","message":"I hardly see a point above of how Rust does solve it comparing to Python.\nThe only potential benefit is performance, which can be potentially addressed with no-GIL and eventlet removal.\n\nSurely, maybe not to the level of Rust, but still decent enough.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":221,"context_line":"Proposed solution"},{"line_number":222,"context_line":"-----------------"},{"line_number":223,"context_line":""},{"line_number":224,"context_line":"The integration of Rust into the OpenStack ecosystem addresses these"},{"line_number":225,"context_line":"structural limitations. Rust delivers near-native execution speed and"},{"line_number":226,"context_line":"deterministic resource control without garbage-collection pauses, while"},{"line_number":227,"context_line":"enforcing memory safety and data-race prevention at compile time through its"},{"line_number":228,"context_line":"ownership model."}],"source_content_type":"text/x-rst","patch_set":3,"id":"4848150e_c2de2b27","line":225,"range":{"start_line":224,"start_character":0,"end_line":225,"end_character":23},"in_reply_to":"b4af82ac_9ef97a7d","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":240,"context_line":"speculative: the Keystone-RS prototype, referenced in the mailing list"},{"line_number":241,"context_line":"discussion [1]_, already implements this approach across more than 170,000"},{"line_number":242,"context_line":"lines of Rust and reports throughput improvements of one to three orders of"},{"line_number":243,"context_line":"magnitude over the current implementation on the benchmarks exercised so far."},{"line_number":244,"context_line":"Embedding Rust into Keystone\u0027s existing Python process via PyO3 was also"},{"line_number":245,"context_line":"raised in that discussion as a possible lower-disruption alternative that"},{"line_number":246,"context_line":"would avoid a service split, but it is not the direction Keystone-RS has"}],"source_content_type":"text/x-rst","patch_set":3,"id":"2b47b173_ef62c88c","line":243,"range":{"start_line":243,"start_character":43,"end_line":243,"end_character":77},"updated":"2026-10-30 06:32:06.000000000","message":"it would be totally prudent to link to these for education and share the benchmarking methodology. I believe this will be your best argument.\n\nI don\u0027t disbelieve that you have made throughput gains. However, I don\u0027t know how much of this is just you fixing the many half-implemented/poorly implemented features in python Keystone (given all the OSSAs/OSSNs we\u0027ve worked on together this year).","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":240,"context_line":"speculative: the Keystone-RS prototype, referenced in the mailing list"},{"line_number":241,"context_line":"discussion [1]_, already implements this approach across more than 170,000"},{"line_number":242,"context_line":"lines of Rust and reports throughput improvements of one to three orders of"},{"line_number":243,"context_line":"magnitude over the current implementation on the benchmarks exercised so far."},{"line_number":244,"context_line":"Embedding Rust into Keystone\u0027s existing Python process via PyO3 was also"},{"line_number":245,"context_line":"raised in that discussion as a possible lower-disruption alternative that"},{"line_number":246,"context_line":"would avoid a service split, but it is not the direction Keystone-RS has"}],"source_content_type":"text/x-rst","patch_set":3,"id":"e5879279_8e942c63","line":243,"range":{"start_line":243,"start_character":43,"end_line":243,"end_character":77},"in_reply_to":"2b47b173_ef62c88c","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"2a49234304a9a4a28142c28e1d935d9bb49afa95","unresolved":true,"context_lines":[{"line_number":281,"context_line":"adoption for Swift. The Keystone team will own CI job maintenance, packaging,"},{"line_number":282,"context_line":"and stable-branch backport policy for any Rust components it introduces,"},{"line_number":283,"context_line":"limiting the initial scope to Keystone rather than requiring ecosystem-wide"},{"line_number":284,"context_line":"tooling changes as a precondition of this Step 1 approval."},{"line_number":285,"context_line":""},{"line_number":286,"context_line":"Conclusion"},{"line_number":287,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"fce33988_18ce28e7","line":284,"updated":"2026-10-20 18:02:05.000000000","message":"This feels shockingly small in the shadow of all the justification above. I think a lot of people have raised concerns that fit into this section, yet it gets two sentences of coverage here. You\u0027ve also said that keystone (-core) is effectively two people at this point, yet seem to act like taking over infra, qa, stable-maint, and release team responsibilities for the new different component tree in addition to the daily tasks seems to be acknowledged and then immediately papered over here.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":276,"context_line":"Community cost"},{"line_number":277,"context_line":"---------------"},{"line_number":278,"context_line":""},{"line_number":279,"context_line":"Adding Rust introduces a second compiled-language toolchain that the infra and"},{"line_number":280,"context_line":"release teams must support, alongside the existing precedent set by Go\u0027s"},{"line_number":281,"context_line":"adoption for Swift. The Keystone team will own CI job maintenance, packaging,"},{"line_number":282,"context_line":"and stable-branch backport policy for any Rust components it introduces,"},{"line_number":283,"context_line":"limiting the initial scope to Keystone rather than requiring ecosystem-wide"},{"line_number":284,"context_line":"tooling changes as a precondition of this Step 1 approval."},{"line_number":285,"context_line":""},{"line_number":286,"context_line":"Conclusion"},{"line_number":287,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"b1f6ebc8_b03c9e71","line":284,"range":{"start_line":279,"start_character":0,"end_line":284,"end_character":58},"updated":"2026-10-20 17:50:14.000000000","message":"This is HUGE undermining of the community cost.\n\n- Community must learn new toolchain and best practises to manage dependencies and requirements management in Rust.\n  - Skyline and nodejs has showcased this is not an easy thing, and since years we have not yet mastered releasing policy for the skyline-console.\n  - This puts anothe constraint to the platform support matrix. Different distributions have different rust versions and dependencies. This has proven to be problematic to align supported rust versions across different distributions.\n- The ask is to add the language for OpenStack in general. This means that it\u0027s not limited to Keystone team who should be maintaining it. It requires to develop approach to the common tools and libraries, alike to oslo, which could be re-used between projects.\n- This requires organization to recruit new developers and personal, who is quite advanced with Rust. And you have rightfully mentioned - identity provider is first class citizen. Thus, organizations require to analyze the source code and be able to debug in case of issues, patch and contribute back.\n- This puts a huge strain on already short in people community by additionally dividing people and stretch thin community even further.\n- It makes pretty much impossible to have a pluggable or extendable desing for such services, which brigs a huge benefit for community and ability to adopt services to their needs, along with possibility for commercialization of products.","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":276,"context_line":"Community cost"},{"line_number":277,"context_line":"---------------"},{"line_number":278,"context_line":""},{"line_number":279,"context_line":"Adding Rust introduces a second compiled-language toolchain that the infra and"},{"line_number":280,"context_line":"release teams must support, alongside the existing precedent set by Go\u0027s"},{"line_number":281,"context_line":"adoption for Swift. The Keystone team will own CI job maintenance, packaging,"},{"line_number":282,"context_line":"and stable-branch backport policy for any Rust components it introduces,"},{"line_number":283,"context_line":"limiting the initial scope to Keystone rather than requiring ecosystem-wide"},{"line_number":284,"context_line":"tooling changes as a precondition of this Step 1 approval."},{"line_number":285,"context_line":""},{"line_number":286,"context_line":"Conclusion"},{"line_number":287,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"c7abc6f3_6bc5a61f","line":284,"range":{"start_line":279,"start_character":0,"end_line":284,"end_character":58},"in_reply_to":"a996853b_0199311e","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":276,"context_line":"Community cost"},{"line_number":277,"context_line":"---------------"},{"line_number":278,"context_line":""},{"line_number":279,"context_line":"Adding Rust introduces a second compiled-language toolchain that the infra and"},{"line_number":280,"context_line":"release teams must support, alongside the existing precedent set by Go\u0027s"},{"line_number":281,"context_line":"adoption for Swift. The Keystone team will own CI job maintenance, packaging,"},{"line_number":282,"context_line":"and stable-branch backport policy for any Rust components it introduces,"},{"line_number":283,"context_line":"limiting the initial scope to Keystone rather than requiring ecosystem-wide"},{"line_number":284,"context_line":"tooling changes as a precondition of this Step 1 approval."},{"line_number":285,"context_line":""},{"line_number":286,"context_line":"Conclusion"},{"line_number":287,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"a996853b_0199311e","line":284,"range":{"start_line":279,"start_character":0,"end_line":284,"end_character":58},"in_reply_to":"b1f6ebc8_b03c9e71","updated":"2026-10-30 06:32:06.000000000","message":"I agree on the two well articulated points above, which have been communicated over various forums as well. \n\ngolang was never officially adopted by any swift deliverable, and we dropped golang as a supported language (https://review.opendev.org/c/openstack/governance/+/989189). A strong reason for dropping it was that we failed to create the mindshare necessary to define the PTI/compiler compatibility policy, release know-how, CI infrastructure (mirrors, common jobs, devstack integration, publishing etc). Rust will need all of this heavylifting. I\u0027m fairly certain it can be done by dedicated volunteers in the community, but right now this looks like a project with a low bus factor. \n\nthis is where I feel using keystone as an example muddies the waters. Right now, we ask the question:\n\nWhile the keystone team takes on the heavy tasks of bringing about rust into the community, who maintains the current keystone? There are established and drive-by contributors writing bugfixes/features but can the small team handle the load to do two massive things at once?","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":281,"context_line":"adoption for Swift. The Keystone team will own CI job maintenance, packaging,"},{"line_number":282,"context_line":"and stable-branch backport policy for any Rust components it introduces,"},{"line_number":283,"context_line":"limiting the initial scope to Keystone rather than requiring ecosystem-wide"},{"line_number":284,"context_line":"tooling changes as a precondition of this Step 1 approval."},{"line_number":285,"context_line":""},{"line_number":286,"context_line":"Conclusion"},{"line_number":287,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"5d00aa22_314742b6","line":284,"in_reply_to":"fce33988_18ce28e7","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":28619,"name":"Dmitriy Rabotyagov","email":"noonedeadpunk@gmail.com","username":"noonedeadpunk"},"change_message_id":"d7235c4fa7affa5ca420831fc054066e86571158","unresolved":true,"context_lines":[{"line_number":288,"context_line":""},{"line_number":289,"context_line":"The OpenStack community believes that the technical and operational arguments"},{"line_number":290,"context_line":"detailed above satisfy the criteria for Step 1 of the language addition"},{"line_number":291,"context_line":"process. The convergence of regulatory pressure, GIL-imposed concurrency"},{"line_number":292,"context_line":"limits, the demand for embedded consensus, a recurring pattern of"},{"line_number":293,"context_line":"authorization-logic and concurrency-safety bugs, and the need to build modern,"},{"line_number":294,"context_line":"crypto-heavy authentication capabilities such as a native OAuth2/OIDC provider"},{"line_number":295,"context_line":"and WebAuthn support makes a case for a language that delivers memory safety,"},{"line_number":296,"context_line":"parallelism, and compile-time enforcement of credential, trust, and"}],"source_content_type":"text/x-rst","patch_set":3,"id":"97a02359_dc5ec3ea","line":293,"range":{"start_line":291,"start_character":9,"end_line":293,"end_character":46},"updated":"2026-10-20 17:50:14.000000000","message":"I\u0027ve addressed all these above...","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":27900,"name":"Artem Goncharov","email":"artem.goncharov@gmail.com","username":"gtema"},"change_message_id":"6d5d052db627e370a61bf6968b1552d6ceb55945","unresolved":false,"context_lines":[{"line_number":288,"context_line":""},{"line_number":289,"context_line":"The OpenStack community believes that the technical and operational arguments"},{"line_number":290,"context_line":"detailed above satisfy the criteria for Step 1 of the language addition"},{"line_number":291,"context_line":"process. The convergence of regulatory pressure, GIL-imposed concurrency"},{"line_number":292,"context_line":"limits, the demand for embedded consensus, a recurring pattern of"},{"line_number":293,"context_line":"authorization-logic and concurrency-safety bugs, and the need to build modern,"},{"line_number":294,"context_line":"crypto-heavy authentication capabilities such as a native OAuth2/OIDC provider"},{"line_number":295,"context_line":"and WebAuthn support makes a case for a language that delivers memory safety,"},{"line_number":296,"context_line":"parallelism, and compile-time enforcement of credential, trust, and"}],"source_content_type":"text/x-rst","patch_set":3,"id":"94726c4b_3656803d","line":293,"range":{"start_line":291,"start_character":9,"end_line":293,"end_character":46},"in_reply_to":"436ef0b4_20fb0de3","updated":"2026-11-06 16:31:04.000000000","message":"Done","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"},{"author":{"_account_id":16643,"name":"Goutham Pacha Ravi","email":"gouthampravi@gmail.com","username":"gouthamr"},"change_message_id":"8cc59fb664df915c3d72ac274ef03c016c9bc547","unresolved":true,"context_lines":[{"line_number":288,"context_line":""},{"line_number":289,"context_line":"The OpenStack community believes that the technical and operational arguments"},{"line_number":290,"context_line":"detailed above satisfy the criteria for Step 1 of the language addition"},{"line_number":291,"context_line":"process. The convergence of regulatory pressure, GIL-imposed concurrency"},{"line_number":292,"context_line":"limits, the demand for embedded consensus, a recurring pattern of"},{"line_number":293,"context_line":"authorization-logic and concurrency-safety bugs, and the need to build modern,"},{"line_number":294,"context_line":"crypto-heavy authentication capabilities such as a native OAuth2/OIDC provider"},{"line_number":295,"context_line":"and WebAuthn support makes a case for a language that delivers memory safety,"},{"line_number":296,"context_line":"parallelism, and compile-time enforcement of credential, trust, and"}],"source_content_type":"text/x-rst","patch_set":3,"id":"436ef0b4_20fb0de3","line":293,"range":{"start_line":291,"start_character":9,"end_line":293,"end_character":46},"in_reply_to":"97a02359_dc5ec3ea","updated":"2026-10-30 06:32:06.000000000","message":"yeah, in this write up, the criteria to meet the Step 1 aren\u0027t met:\n\n- Python 3.13+ has no-GIL - is this insufficient too?\n- Has the team experimented with modern async python patterns  \n- Community cost grossly understated\n- No consensus demonstrated","commit_id":"d534c6dfa2da48083a7a717a3ec9ec2647951a36"}]}
