<gchristensen>
with a detour in letters, sorts, upper and lower cases, etc.
<__monty__>
Does it apply if I'm not a rubyist?
<gchristensen>
yes
kini has joined #nixos-chat
<sphalerite>
gchristensen: I mean why we have two versions of every letter, not why they're named uppercase and lowercase :p
<gchristensen>
oh, well, watch the talk anyway :P
<jasongrossman>
I asked my wife why we have upper- and lower-case letters. She's a linguist who knows everything about language.
<jasongrossman>
She doesn't know.
<__monty__>
Does it maybe come from the fancy letters at the start of chapters?
<gchristensen>
if I had to guess it is an affordance so it is easier to differentiate
kini has quit [Quit: No Ping reply in 180 seconds.]
kini has joined #nixos-chat
<jasongrossman>
It's very old ... I'm not sure whether it's older than manuscripts.
kini has quit [Quit: No Ping reply in 180 seconds.]
pie_ has quit [Remote host closed the connection]
pie_ has joined #nixos-chat
kini has joined #nixos-chat
spacekookie has joined #nixos-chat
<infinisil>
Idea: In NixOS-land we can do fancy rollbacks, but these only work if you didn't change things like the filesystem inbetween. So how about we split generations into two: One generation timeline for the hardware-independent config, and one for the hardware-dependent one. Then you could rollback each of them separately
<infinisil>
Probably can't do because those parts are interdependent..
<joepie91>
gchristensen: excellent talk
<manveru>
gchristensen: thanks man, i watched this years ago, totally forgot about it :)
<pie__>
tl;dr: code separated into pieces is nice, now if only it was possible to override individual pieces
endformationage has joined #nixos-chat
drakonis1 has joined #nixos-chat
kgz has quit [Quit: WeeChat 2.3]
kragniz has joined #nixos-chat
kragniz is now known as kgz
waleee has joined #nixos-chat
drakonis1 has quit [Quit: WeeChat 2.4]
drakonis_ has joined #nixos-chat
yorick has joined #nixos-chat
drakonis has quit [Ping timeout: 246 seconds]
<zimbatm>
yorick: how much time did you spend on this already?
<yorick>
zimbatm: time tracker is currently at 1 month fulltime
<zimbatm>
roberth should be working on tree-sitter-nix binding for Haskell as we speak
<zimbatm>
oh wow
<yorick>
it does all of nixpkgs in 20s
<yorick>
(given 4 cores)
<zimbatm>
I suppose your design means that the formatter is also enforcing line lenghts isn't it?
<yorick>
yes, that's the only parameter
<yorick>
(but that means you can do real-time nix reflow :D)
<zimbatm>
heh :)
<yorick>
I'm currently compiling ghcjs for a web demo
drakonis_ has quit [Ping timeout: 252 seconds]
<pie__>
yorick, what are you working on
<pie__>
meanwhile...can i use nix-copy-closure with my current ssh key somehow?
<yorick>
pie__: serokell is working on nix formatter
<pie__>
or can i somehow pack, transfer, and unpack a derivation ? nix-store --export doesnt look like the right thing
<yorick>
pie__: yes use ssh-ng://
<pie__>
yorick, aha
<yorick>
pie__: um, you can nix copy --to ./asdf, then targz asdf, then untargz it on the other side, then nix copy --from ./asdf
<pie__>
ah ok
<pie__>
where do i find info about this ssh-ng thing
<yorick>
the manual maybe
<yorick>
there's a whole line on it SSHStore (ssh-ng://) supports arbitrary Nix operations on a remote machine via the same protocol used by nix-daemon.
<pie__>
error: don't know how to open Nix store './plz'
<yorick>
file://./plz then
<pie__>
ok file:// worked
<pie__>
yorick, very verbose. I mean can i just pass -i or something? i dont want to use a .ssh/config
<__monty__>
Is it just because you all have OCD? : >
<gchristensen>
for me, I do it because why have state if I don't need to
<etu>
Same here, it's much easier to get an overview of what data is actually stored on disk etc
<etu>
if you stop using some software, you can reboot and all traces are gone.
<aanderse>
yeah i like the idea of very explicitly stating mutable data
waleee has quit [Quit: WeeChat 2.4]
pie__ has joined #nixos-chat
<pie__>
waiting for the day someone says "we have the dependencies, we can rebuild hin
<pie__>
*him
<MichaelRaskin>
I have a feeling that keep-derivations + keep-outputs means that different versions of curl accumulate
<MichaelRaskin>
I also have a feeling that for a set of packages from binary cache the build closure is more than two times larger than runtime closure
<MichaelRaskin>
… which makes the accumulation of fetchurls less relevant, of course
<samueldr>
we have the dependencies, we can rebuild bin
<__monty__>
Anyone know if there's a name for the hexadecimal floating point notation specified in ieee-754? Does it come from an SI convention or something?
<joepie91>
yeah, this looks like private registry hosting
<andi->
soon you wont be able to write programs that require third party libraries without having Microsoft know about it? :-)
<joepie91>
hardly
<joepie91>
if this is what I think it is, it's basically hosted Verdaccio/Artifactory (from an npm perspective anyway)
<andi->
ofc it isn't enforced but it is super convenient (nothing wrong about that)
<samueldr>
and more to the point: feature parity with gitlab
<samueldr>
though looks like github's will support more kind of registries
<samueldr>
gitlab seems to only have docker, maven and npm
<joepie91>
I can't help but feel that this announcement is going to elicit some drama
<joepie91>
in the longer term
joepie91[m] has left #nixos-chat ["User left"]
jasongrossman has quit [Quit: ERC (IRC client for Emacs 26.2)]
<__monty__>
Why?
<joepie91>
__monty__: because this is quite literally eating npm's business model
<joepie91>
I don't imagine that npm inc is going to be very happy about that
<_e>
It seems simalar to what aws has done to mongo and elastic search.
<joepie91>
and this could well result in an ecosystem breakdown
<__monty__>
joepie91: npm doesn't exactly have a respectable track record though. So maybe it's not a big loss?
<joepie91>
__monty__: with an ecosystem breakdown I don't mean that npm gets replaced
<joepie91>
I mean that shit actually starts breaking
<joepie91>
which is definitely worse than the current situation
<joepie91>
I might find that an acceptable risk for an overtaking by a community/foundation-led registry
<joepie91>
but for moving to another random commercial provider? no thanks
<__monty__>
You mean JS disappears off the face of the planet? One can only hope : >
<gchristensen>
ouch
<joepie91>
welp, guess that marks the end of the conversation.
<gchristensen>
not cool, __monty__ :(
<joepie91>
would be nice if people could, for once, not do this shit
<__monty__>
What's a language joke among friends?
<gchristensen>
it is hurtful to be on the receiving end of such a joke
<joepie91>
it's really not funny or witty
<joepie91>
it's the punchline to every goddamn conversation I try to have about JS outside of JS channels
<joepie91>
and I am /really/ tired of it
<joepie91>
hundreds have made this 'joke' before you
<joepie91>
if you can even call it a 'joke'
<simpson>
Yeah, it really poisons the well for those of us working on alternative languages to JS.
<__monty__>
Apologies. It's a deserved reputation afaics. Didn't mean anything by it.
<gchristensen>
I mean
<joepie91>
no, it's very much not.
<gchristensen>
your apology says you did mean it
<simpson>
joepie91: FWIW the way that I see this move by Github is that we will now have *two* large commercial providers of Node-compatible packages. Perhaps that will lead to even more than two. And perhaps that will, in turn, lead to a more flexible ecosystem.
<joepie91>
simpson: unfortunately, I doubt that, because the entire dependency architecture is built for having *one* central registry.
<joepie91>
and npm owns that namespace.
<joepie91>
the best-case scenario here is that npm inc works together with github on not having name clashes.
* colemickens
tangentially wonders how rust is going to handle alternative non-Crates.io registries.
<joepie91>
but I sincerely doubt that that will happen.
<joepie91>
the more likely result is that there will come to exist two parallel dependency ecosystems that don't interoperate
<joepie91>
or that the github registry just doesn't get used for public packages because people want to avoid exactly that
<joepie91>
this whole npm-owns-the-namespace problem is one I've been working on for a while, and that recently somebody brought up a potential solution for
<joepie91>
but it seems that github ate my lunch
<simpson>
Or people write a new package management tool which puts both npm and github into their own namespaces.
<joepie91>
and just barged in and said "fuck it, we'll run it in parallel"
<joepie91>
simpson: there are significant backwards compatibility issues with that.
<joepie91>
it is /technically possible/ to work around this problem, most likely, but I see no evidence of github attempting to do so
<joepie91>
and it involves a nasty hack with fake Git servers
Synthetica has quit [Quit: Connection closed for inactivity]
<simpson>
joepie91: If that's the case, then it's a tacit admission that folks have been writing code to npm, and not to Node or any more-open governing body. It's always unpleasant to discover one's underlying corporate sponsor.
<joepie91>
this has been known for years
<joepie91>
it's a network effect problem
<__monty__>
gchristensen: I didn't mean to insult anyone. I *did* mean npm's reputation is terrible when it comes to security and JS's for violating the law of demeter. I'm not saying other package ecosystems have better security practices btw, just that npm *has* a bad reputation.
<joepie91>
npm != JS
<gchristensen>
samueldr: how do I make a pre word wrap? :/
<samueldr>
depends, does it need to break a word?
<gchristensen>
no
<samueldr>
word-wrap: normal; I think should do it
<gchristensen>
hrm
<samueldr>
and breaking a word is almost always going to cause pains in the end
<gchristensen>
eh, I won't worry about it
<samueldr>
wait, white-space: pre-wrap, word-wrap is how the word themselves are going to be wrapped
<samueldr>
going to be broken*
<__monty__>
simpson: How does a joke about JS poison the well for alternatives? Wouldn't the demise of JS benefit replacements rather?
endformationage has quit [Ping timeout: 246 seconds]
drakonis has joined #nixos-chat
drakonis_ has joined #nixos-chat
__monty__ has quit [Quit: leaving]
drakonis has quit [Quit: Leaving]
drakonis_ has quit [Ping timeout: 258 seconds]
<andi->
One more day and my systemd v242 branch is down to a single vm test failing (that doesn't fail on hydra.nixos.org) \o/ I hope that by sunday I can create all the PRs… We should do that more often and not just once a year :/
<gchristensen>
nice!
<andi->
(side note; digital ocean S3 buckets are really slow when inserting.. not sure if I'd do that again)
<andi->
and the 250GB cap doesn't help either.. will have to wipe it once done..
<gchristensen>
nice
<samueldr>
don't know if it can be useful, but nix can also insert into a cache over http; I have a barebones thing I used to create a small cache
<andi->
Will probably go that path in the long run. Did the setup a while ago and it worked really well so far. INserting large amounts of files just became a bit slow and hydra would sit there downloading form cache.nixos.org and reuploading to my cache for hours initially..