httpxyz one month in
26 points by knl
26 points by knl
One of the bullet points here is that some commits were reinstated into httpxyz which were removed from httpx due to “policy”. Well the commit in question has a big fat co-authored by Claude on it. Not passing judgement, but there’s clearly a lot going on here.
The original PR wasn't written by Claude: https://github.com/encode/httpcore/commit/973cbdd1050302822f17893c951ca5a1a729dc13 (though I don't think httpx has any policies against AI either?)
I also think the commit isn't amazing because it's working around an event loop problem -- a program being slow due to too many yields to the event loop is a good thing! much lower tail latencies! -- and should be fixed there, e.g. by making anyio.lowlevel.checkpoint() not yield to the event loop until a timeslice is over. Unfortunately nobody has done the necessary optimization for Trio or asyncio yet.
Yeah, I did some digging by going through the list on "Why I forked httpx" article.
It glosses over a lot. Firstly, httpx is written by a trans/queer person — Mia Kimberly Christie. The examples of repositories that are dependent on httpx that might be broken by a fix, are all AI companies that are depending on the unpaid, unfunded work of a queer person.
The link to Simon Wilson's comment honestly looks like a decent proportion of the writing at that link was vibe-written, and the final link "The Slow Collapse of MkDocs" is a vibe-coded website with what looks like an entirely vibe-written analysis of the history behind Mia's project, MkDocs.
I spent the most time reading the MkDocs article — it's trying to present an argument that calls into question the reliability of the Mia, but if you look between the lines it's just a flat out character assassination. It glosses over any kind of transphobia or hostile speech that Mia has had to deal with, as well as Mia's calls to move discussion into a place that has actual moderation that is not a collection of github issues where they can be consistently shouted over and brigaded, while also trying to put a positive, almost righteous spin on a very blatant hostile takeover of their repositories.
It's icky, and the fact that this author linked to it (that is, either they are ok with the article misrepresenting hostile speech, adversarial behaviour, and transphobia, or they never dug deeper) is icky, and the whole thing smells rank from the bottom up.
As a queer person, yeah I completely fucking understand rug-pulling any kind of github comment section, if those are the sorts of comments I could look forward to when I open github. All the actions that Mia has taken here feel entirely reasonable under the actual circumstances, quite frankly.
First, I fully sympathize with Mia, but:
The examples of repositories that are dependent on httpx that might be broken by a fix, are all AI companies that are depending on the unpaid, unfunded work of a queer person
No? Any combination of libraries that depends on httpx will break. Simon used those two because that's a combination widely used in practice. There's hundreds of thousands of packages that depend on httpx.
I also disagree with the "evil AI companies are selfishly profiting off her work" narrative. Both OpenAI and Anthropic are more than willing to fund open source and send patches for bugs. The problem at hand is that Mia is refusing to merge bugfix PRs and tag a new non-v1 release. Notably, they don't need to be the one reviewing or merging, so long as they're willing to share the maintenance burden. For a package that's become so foundational to so many python packages, I think that's a fair request to make.
The link to Simon Wilson's comment honestly looks like a decent proportion of the writing at that link was vibe-written
Simon has said multiple times in the past that he doesn't use LLMs to write text for him, and all the text in that post matches his regular style - this is an odd assertion.
Regardless, I feel like that's missing the point that OP tried to make here. I can fully understand checking out of a community because it is toxic and transphobic, but when millions of people across the world (and not just the big evil AI companies) are stuck waiting for one person to at least tag a patch release, and that person refuses to do so and insists that everything has to wait for a breaking, incompatible v1 release, I don't think that's good or defensible.
Both OpenAI and Anthropic are more than willing to fund open source and send patches for bugs.
lol. On the httpxnext page (which is on the same page as the discussion with Simon Wilson, so, you had the chance to spot it) it shows that Mia's organisation Encode is looking for funding.
We are currently seeking forward-looking investment that recognises the value of the infrastructure development on it's own merit. Sponsorships may be made through GitHub..
Looking at the sponsors present, it's pretty clear that while a number of AI companies have funded their company for work (AI companies love logos that look like an anus), and a handful are currently funding them, none of the sponsors come from the major-league companies who are profiting off their work (as mentioned- Anthropic, OpenAI), and honestly I really doubt that they're pulling enough to make it worth working on the project versus doing any other profitable endeavour.
The problem at hand is that Mia is refusing to merge bugfix PRs and tag a new non-v1 release. Notably, they don't need to be the one reviewing or merging, so long as they're willing to share the maintenance burden.
The reason why I state that the article is a blatant character assassination is because if you actually read the discussion,
Yep the project could totally have a different PyPI name. Getting that right is the thing ain't it?
Mia has been very consistently open about accepting advice and input from people. Right up until the point where someone tried to do a hostile takeover without talking to them, and then in the same discussion, they got deadnamed. I think that the former represents such an enormous breach of trust that it is an entirely fair and justifiable reason for pulling the drawbridge up and not trusting possible future maintainers. I would do exactly the same, quite frankly.
but when millions of people across the world (and not just the big evil AI companies) are stuck waiting for one person to at least tag a patch release, and that person refuses to do so and insists that everything has to wait for a breaking, incompatible v1 release, I don't think that's good or defensible.
Maintainers do not owe anyone else their work. Especially not in the face of a lack of funding, brigading, bullying, character assassination, and hostile takeovers. Moreover, based on the last year that trans folks worldwide have had, but especially in the UK and the USA, and the sheer amount of bullshit we have had to deal with, I think that cis people deserve even less from trans people as a whole. I think that you're quite lucky that they've even bothered to keep the repository up in the first place.
First off, I think you're being aggressive to me for no reason. I would hope that you approach discussions here with at least a presumption of good faith.
The reason why I state that the article is a blatant character assassination is because if you actually read the discussion,
I'm not talking about the mkdcos article or that situation. I'm referring to the point OP made in this post, which is just what I said and nothing more: Mia is refusing to merge bugfix PRs and tag a new non-breaking v1 release, which a singificant portion of the httpx userbase does not want. As a reference, this is literally what happened with mkdocs and that is what OP cites. I fully agree that the discussion there is toxic, but I don't see OP justifying that.
Maintainers do not owe anyone else their work
I don't think this line applies when your package has become a foundational part of the entire ecosystem. At that point, it's not a hobby, and you should consider better governance and letting go of control, especially if how you want to steer it is clearly different from what the literal millions of users want.
I also disagree with the "evil AI companies are selfishly profiting off her work" narrative. Both OpenAI and Anthropic are more than willing to fund open source and send patches for bugs.
I would hope that you approach discussions here with at least a presumption of good faith. I fully agree that the discussion there is toxic, but I don't see OP justifying that.
I mean, you dropped the point about Anthropic et al profiting off their work simply because I had disproven you on that matter, right? And you really can't see how the MkDocs situation ties into what is essentially a very hostile work environment for the lead maintainer? If the person's work is a "foundational part of the ecosystem", then uh, maybe they should have been given a larger amount of funding by that ecosystem, and perhaps that ecosystem should have treated them better.
At that point, it's not a hobby, and you should consider better governance and letting go of control, especially if how you want to steer it is clearly different from what the literal millions of users want.
Nah, what I produce is my art. I build a pillar, and other people decide, irrespective of what I want or not, to build their multibillion dollar industries upon my pillar, and moreover do not pay me back even a portion of that, it is kind of their fault then if I get tired of being shouted at over that pillar and decided to tear it down in the night for a different looking pillar that they happen to dislike.
I think that for two decades the OSS movement has allowed companies to profit relentlessly off of people who make software without the expectation of coughing up any kind of recompense, while also having disproportionate say in the direction of those projects, in a very "capitalism for me, community compensation and Exposure for thee" way, and at some point that falls down.
Either you compensate people for their time, and treat them fairly, or you reap the exact rewards of what you sow. We live under capitalism, and until we stop living under capitalism, maintainers still need to feed and clothe and house themselves. Trans people still need healthcare and surgery and stability and to stop being abused by the systems they live under. Trans women earn 50-60 cents on the dollar, with trans men on average earning 70-80 cents on the dollar and cis women earning 80 cents on the dollar. So you're expected to outperform your peers just to get respect, for ½ the wage that they're getting, often getting talked down to in the process, and then you go home and people talk down to you about your spare time projects?! What?? It especially falls down when you're living in a situation where, when you wake up and turn on the news, you find another round on people are debating on if you're allowed to use public bathrooms or not. Another round of people debating if you're human, etc. And then you go to work and get met with ??? more bullshit? More entitled users who have built things on your vision and not given you anything in return?
Nah. The people involved should have done better.
I think I was originally mentioned in this comment (and I couldn't figure out why I received a notification until I read the article). Just to make things perfectly clear, especially since the discussion is getting heated: I am not the author of this article. I am a different person with the same (fairly common) first name who happens to have an account on lobste.rs.
Weird, I don't remember mentioning anyone! (I am reasonably sure i didn't use an @ sign in the entire thing). Sorry if that's the case though, uh.
I appreciate being the change they wish to see in the world migrating to Codeberg… which has seen a small uptick in new users if their report at the end is to believed. The big tech corpos have enough control as it is—no need to add to their centralization of power.