What's cooking on SourceHut? Q2 2025
55 points by ThinkChaos
55 points by ThinkChaos
A small improvement to the GraphQL APIs includes the addition of better standardized errors, so that you can distinguish error conditions on the client side – for example, readily distinguishing “not found” from “access denied” in your GraphQL client.
I remain flabbergasted that this completely solved problem remains a thing that you have to do extra work to get with GraphQL instead of getting it for free.
At work we recently finally got our stuff off GraphQL years after engineering management made us go all-in on it and it’s been such a relief.
It’s really too bad that we never invented a proper micropayments system for the internet. The problem of AI scrapers would not be a problem if they’d just have to pay for every bit they ingest.
It’d be a different Internet for sure, but I’m sure some smart person could have figured out a way to make it work without killing the soul of the thing. And of course many other monetization methods have been invented that totally killed it.
I’m pretty sure they would find a way to launder data retrieved through botnets that paid with someone else’s money…
You’d be surprised. I worked in VoIP business for quite a while. Whenever someone offered you a “route that didn’t support caller ID”, it was implicitly understood that the route went through something like a box with a thousand SIM cards in it. Often extracted from phones to be unlocked for gray import — they have $5 on prepaid SIM cards that come with them, so why not put them to use? The legit company that offered such routes for resale could buy them from a sacrificial company registered in some place with really loose laws or spotty enforcement, and tracking down the location of a SIM card that isn’t moving is only possible up to its cell.
There were manufacturers that produced GSM gateways that could load SIM data from a remote source. As in, in one place you have a box that holds a thousand SIM cards. Then you have gateways that load the SIM card data over IP. Can you see a legitimate use for such hardware, other than to make such schemes safer to operate — the place that has material evidence (physical SIM cards) doesn’t emit any GSM signal, and the place that emits signals has no material evidence.
After I’d seen those “non-CLI routes”, I think a “market” for pushing the cost of micropayments onto someone else would arise quickly.
It used to be that you would see a lot of blogs with Bitcoin and Dogecoin addresses back in 2014, so that people would send a few cents if they wanted to.
Micropayments as part of the web platform can’t really work without dramatically reducing the scope of the platform, probably effectively to the US only (and maybe a smattering of other rich countries). So I’m not really convinced it would even be the internet any more.
the success of Anubis and go-away in mitigating the LLM bots problem
I love to hear success stories like this, especially ones where it remains a success when moving from Anubis to go-away.
I know that LLM abuse derailed the roadmap, but I would like to know know at least what are the plans. But I am curious what are they planing regarding
improvements to the lists.sr.ht code review workflows
which Drew mentioned on ther their What’s Cooking September 2024. imho that is the weakest part of sourcehut and also have the biggest learning/onboarding curve.
As a sidenote, although I prefer sourcehut, it seems that forgejo has been more successful in building a larger community around the project. I do wonder why has that happened.
it seems that forgejo has been more successful in building a larger community around the project. I do wonder why has that happened.
forgojo’s lineage is older (its a fork of Gitea that itself is a fork of Gogs), and those all are much more providing an experience similar to other forges, whereas sourcehut is very intentionally different, and the audience that loves that is seemingly a lot smaller.
But it’s indeed kind of interesting that I can’t think of a single person or project hosting it themselves. Is or was it much harder to host? Do the kind of people who like it don’t see the need and go with even more basic setups if they self-host?
Is or was it much harder to host?
When I tried it a few years ago (before forgejo forked), Gitea took about 10 minutes to set up if you want to get your own personal instance going. I haven’t looked into doing the same with sourcehut, but I do know it’s made up of several different services, some of them written in python. So I’d expect it to be an order of magnitude more work to run your own, at minimum. Probably significantly more.
I still don’t know why Drew didn’t add a billing integration with Liberapay. It feels like it’s an option that’s better ethically aligned with his and Sourcehut’s principles.
I would assume it is because LiberaPay is as far as I know not for purchases, but for recurring donations. Which is not really what SourceHut needs. Also it seems that LiberaPay still uses Paypal and Stripe.
To elaborate, Liberapay can basically only exist in it’s current form because they ban any sort of reward for the donation. The reason is that a lot of finance and business regulations start appearing once you start offering services for money. LiberaPay lets you donate to FOSS maintainers from anywhere without charging any fees for the same reason that you can give money to a busker without them needing to register a legal entity, charge you VAT, hold onto your payment for a 14 day refund window, display their business address and a support phone number, or concern themselves with claims against an implied warranty on their busking services.
In addition, I think what also plays a role is that LiberaPay has very few users and thus probably would cost more to support than it would bring benefits.