andi- has quit [Remote host closed the connection]
greizgh has quit [Quit: greizgh]
greizgh has joined #nixos-dev
andi- has joined #nixos-dev
xwvvvvwx has joined #nixos-dev
xwvvvvwx- has joined #nixos-dev
xwvvvvwx has quit [Ping timeout: 240 seconds]
drakonis1 has quit [Quit: WeeChat 2.6]
alp has left #nixos-dev ["Leaving"]
alp has joined #nixos-dev
{^_^} has quit [Remote host closed the connection]
{^_^} has joined #nixos-dev
<andi->
What is the reasoning behind the custom base-32 in Nix? The characters `E O U T` are emitted. I could guess that it is relic and at some point Nix did scan for `out` or such?
ixxie has joined #nixos-dev
<Taneb>
andi-: apparently it was to minimize the chance of swear words appearing in the hash
<eyJhb>
What, the, hell. Does that matter, at all?
<andi->
Taneb: source for that? :D
<Taneb>
andi-: I thought I read it in Eelco's PhD thesis but I may be mistaken
<andi->
ahh, worth checking that again
<andi->
Taneb: you are indeed right: This is to reduce the possibilitythat hash representations contain character sequences that are potentially offensive to someusers (a known possibility with alphanumeric representations of numbers [11]).
<qyliss>
domenkozar[m]: AAUI, as soon as 5.4 comes out, 5.3 is considered EOL
<qyliss>
I don't think there's really an upstream flow for what you want to do, and we can't reasonably support kernels that aren't supported upstream.
<samueldr>
looks like any version is EOL'd in about the month following the release of the next version
<domenkozar[m]>
qyliss: consider two scenarios
<domenkozar[m]>
when 5.3 is EOL
<domenkozar[m]>
1. I get an error: kernel 5.3 is end of life, pick another kernel
<domenkozar[m]>
I type kernel.org, pick one
<domenkozar[m]>
fix nix, rebuild
<domenkozar[m]>
1min of work tops
<tilpner>
domenkozar[m]: You can put an assert into your config that will error if it's not using 5.3
<domenkozar[m]>
2. I'm forced to use _latest, which I don't notice during channel bump, and 4 weeks later I reboot because my machine freezes. I boot and it hangs. I lose 4h debugging what happened
<domenkozar[m]>
why are we giving up evaluation errors for users time?
<domenkozar[m]>
tilpner: how is that different to having _5_3 attribute?
<tilpner>
Now your config breaks if you update to a nixpkgs where _latest is not 5.3
<tilpner>
No accidental upgrades
<qyliss>
domenkozar[m]: because we shouldn't be encouraging users to depend on specific non-LTS kernel versions.
<domenkozar[m]>
tilpner: that's perfect, let's have it upstream :D
<domenkozar[m]>
qyliss: I understand the quest for stable api, but that's just not always the best path imho
<domenkozar[m]>
on top of that, we can instruct what to do when api breaks
<domenkozar[m]>
I do think that's important
<qyliss>
Consider this scenario: one of the nice things about stable is you can set up automatic upgrades on your server and be reasonably confident it will keep automatically updating forever
<qyliss>
If you pin to 5_3, that breaks
<domenkozar[m]>
you can be reasonably comfortable it's going to break
<gchristensen>
yeah, the "stable" branches get a nearly random collection of patches. whatever happens to be selected for backport by the AI-powered backport bot.
<gchristensen>
(for those playing buzzword bingo in my private messages, yes this AI powered backport bot is also blockchain since it is writing to git)
<qyliss>
What does it take for "Nix" to reach buzzword status?
<cransom>
i suspect it would be some large entity publicly advertising they are going to throw a big chunk of money at it with some press release with a clickbaity title like 'chef/ansible/salt are dead.'