IRC channel logs

<rekado>discussion about nixos: https://news.ycombinator.com/item?id=30057287

<rekado>it seems that now people keep mentioning guix on nixos discussions. Maybe this is payback for “mandatory mention of nix” in Guix discussions… :)

<drakonis>ha

<drakonis>yes.

<drakonis>guix comes up but it isn't always being spoken of kindly.

<drakonis>you even have the classic "why not common lisp"

<drakonis> https://news.ycombinator.com/item?id=30060160

<drakonis>its not exactly a very good argument in favour of rewriting it in CL

<civodul>rekado: and a great reply of yours!

<civodul>"The documentation is gigantic and it's hard to find what you're looking for." uhuh

<drakonis>that's a first

<drakonis>i've never seen anyone complain about the documentation

<drakonis>y'know, why does it always feel like guile is the closest thing to CL in scheme?

<drakonis>the only thing it is missing is in the interactive development side of things

<drakonis>the emphasis on operating on a running application doesnt appear to be as strong yet

<drakonis>ah i'm rambling

<drakonis>i have a plan for the nix daemon rewrite after it reaches parity with the nix daemon

<drakonis>the main takeaway is that CL's development workflow is very good

<rekado>we’d all benefit from improved debugging features in Guile

<rekado>I don’t have any practical CL experience, so I don’t have a good feeling fo what I’m missing

<rekado>I think, though, that too often people imagine contributing to fix annoyances to be too difficult.

<civodul>the good thing is that i think we receive more non-package contributions than Nix (proportionally)

<drakonis>good lord, nixpkgs' stats are out of control

<drakonis>3k open PRs and more than 5k issues

<drakonis>and the majority of open PRs are in the last two months

<drakonis>the churn is real.

<drakonis>too much upstream and forget here

<civodul>heh, and i thought our own stats were out of control :-)

<drakonis>ha, far from it!

<rekado>maybe that’s where we’re headed though?

<drakonis>probably, but at the current pace, it'll be in half a decade?

<rekado>isn’t that about the head start that Nix has over Guix…?

<drakonis>less if guix ramps up and everyone wants to upstream one shot packages

<drakonis>nix has a decade headstart iirc

<drakonis>the cause of the problem is that everyone wants to upstream their leaf packages and that causes a massive inflation in stats

<drakonis>what also contributes to the problem is that they have a problem where adding more repos is very difficult

<drakonis>there's still no reasonable way to do multiple channels there

<drakonis>so they couldnt do a hypothetical setup like a channel where the package churn and updates goes first

<civodul>sometimes i wonder if we really want to keep growing indefinitely

<drakonis>its a social problem with them

<drakonis>they're too far in to actually institute such changes

<drakonis>the approach does not scale well because of the emphasis on the monorepo and passing the responsibilities to the nixpkgs maintainers

<drakonis>they've been trying to offset it with approaches like teams and giving package maintainers free reign over their packages

<drakonis>it is a fine case to learn from

<civodul>i guess you can't have cohesion at this scale, socially speaking

<civodul>Debian manages, but it's probably smaller

<drakonis>their setup does not lend well to it anyways

<drakonis>debian has as many packages as nix does if you exclude the generated package sets

<drakonis>which does nothing but inflate their numbers

<drakonis>honestly though, i'd rather form an ecosystem of channels instead of trying to infinitely inflate the package count

<civodul>yes, but the difficulty is finding a boundary

<civodul>like, you'd say "let's have a Haskell package"

<civodul>but the next day, glibc depends on pandoc

<civodul>s/package/channel/

<drakonis>then you'd have a strong base in which the channels can rely on

<drakonis>you'd need to have such a base.

<drakonis>the boundary would be package that aren't someone's niche one time package

<drakonis>good lord github is taking forever to open the issues and PRs now

<drakonis>in nixpkgs

<drakonis>i do agree that finding a boundary is very difficult because not everyone might agree on where it should be

<drakonis> https://github.com/NixOS/nixpkgs/pulls?q=is%3Apr+is%3Aopen+label%3A%228.has%3A+package+%28new%29%22

<drakonis>regarding guix's growth, right now we're at the point where nix was in 2016?

<drakonis>we have the channel regulars and the increasing influx of new people

<drakonis>civodul: in your opinion, does shepherd's current state warrant an overhaul, a rewrite or a replacement?

<civodul>drakonis: i think it requires work :-)

<civodul>but not necessarily throwing it all away

<civodul>we need to add a proper event loop first, and from there we can build more interesting things

<civodul>i think it can remain rather small

<drakonis>ah, an event loop would be very desirable.

<drakonis>i'm perhaps a bit too prone to scope creep

IRC channel logs

2022-01-24.log