- 55comments
- 49comments
- 73comments
- 28comments
- 593comments
- 136comments
- 6comments
- 230comments
- 83comments
- 170comments
- 72comments
- 41comments
- 234comments
- 2comments
- —discuss
- 158comments
- 12comments
- 117comments
- 24comments
- 59comments
- 29comments
- —discuss
- 45comments
- 110comments
- 371comments
- 12comments
- 39comments
- 60comments
- 49comments
- 39comments
It's a single WRT54g running OpenWRT on an admins DSL connection.
I thought it ran on a 10 year old smart fridge.
Some recent discussion: https://news.ycombinator.com/item?id=27452276
Thanks, this didn’t show up when I searched. Still no info about the architecture there, but Lisp still seems to be the man of the show.
Tangent:
An (early?!) version of HN source code is available on GitHub [1]. It comes with a custom LISP dialect, ARC, that is itself running under MzScheme.
This architecture looks pretty arcane. Has anyone an idea what lead Paul, et. al. to write a custom LISP dialect for building a web forum? Is that a common thing for schemer's to do to get things done? Or does the ARC language predate the HN implementation? I am not sure.
[1] https://github.com/wting/hackernews
I think I saw in one of his blogs that he used arc elsewhere. I imagine "arc.arc" and "html.arc" were relevant to other projects as well, so I doubt they were developed just for this.
Someone can correct me if I'm wrong but AFAIK Hacker News was meant to be a MVP for the Arc language itself.
Also a more up to date version of the Arc forum can be found at https://arclanguage.org, and there is a public fork, Anarki at https://github.com/arclanguage/anarki which is not in feature parity with the former, which itself is not in feature parity with Hacker News.
This knowledge is too powerful, no one can know!
Dan himself takes an amnesia pill whenever the project needs to be modified.
Jokes asides, I'm also very interested in knowing the infra of HN. Comments take a few seconds/minutes to actually appear so I'm guessing there's some queueing involved and probably a lot of caching on the articles?
It's still the same as what Scott described here:
https://news.ycombinator.com/item?id=16076041
Number of daily requests has gone up closer to 6M
So... No microservices SPAs with noSQL database on top of kubernetes in a monorepo?
So disappointing.
No caching?
Interesting.
We use an Nginx front end for that. It all runs on the same box though.
Caching on the same box as everything else is running?
I'm guessing the server must have a ton of RAM, but that's probably why the site is super fast.
AWS lambdas
It runs on a quantum AI blockchain and has recently been rewritten in rust /s
Before I read up on it, let me guess: PHP or Perl.
Update: Oh damn. I took the primitive vibe as a sign the technology would be too. And well, it’s written in a language pg invented.
Other than that it does indeed seem very primitive, though (hosted on some Xeon FreeBSD server, for instance). The only (slight) magic seems to be in ranking/karma.