<andi->
Just remebered that we are in 2020 and 2019 was supposed to have some work on Package managers + IPFS (including Nix).. Sadly there is no more word on the topic on their roadmap :/ https://github.com/ipfs/roadmap/blob/master/2019-IPFS-Project-Roadmap.md It is a tough problem.. Would have been too good
<worldofpeace>
the data I see from the api is empty though
<gchristensen>
right, I wouldn't call this a "shadow" ban
<gchristensen>
they're pretty much just gone
<samueldr>
yeah, let's not dwell on the misuse of "shadow ban" (which most often means they don't appear banned and are just cut off from interacting publicly)
<gchristensen>
right on, fair enoug
<samueldr>
something's up and all their info, including past contributions seem... vanished
<gchristensen>
shall we -> -dev?
<samueldr>
probably
<gchristensen>
maybe github is having a DB outage
<gchristensen>
oops
<gchristensen>
45 minutes and 66% in to netbooting this server with apparently horrible basic-level firmware for its NICs
<gchristensen>
10Gbps NIC, but can't manage to download a 400MB ramdisk in less than an hour
<andi->
TFTP?
<gchristensen>
iPXE and http!
<andi->
The worst I have seen was 40 bytes a packet and then the next TCP roundtrip would get another 40 bytes..
<gchristensen>
let us hope that this installer works first try.
<andi->
I do not understand how companies that produces these crappy broken devices can stay in business.. Every single engineer that I talk to wouldn't accept that as the level of quality for their product :/ Maybe that is just the social bubble all over again
<gchristensen>
andi-: I did a bad job of twittering there. I should have put the context of the issue in to that reply with the ticket #
<joepie91>
andi-: there are plenty of developers who either don't care or, more commonly, do not feel in a position to question technical decisions
<joepie91>
I've lost count of how often I've explained to someone "approach X is a bad idea because <technical reasons>" only to get back a response of "yeah well it's not up to me"
<andi->
joepie91: yeah, the reality is cruel..
<joepie91>
(after they asked for help with approach X)
<andi->
gchristensen: rephrase & retweet ? ;-)
KeiraT has joined #nixos-chat
<joepie91>
andi-: then when I spend 15-30 minutes explaining to them that as a developer, it is *exactly* their position to question technical decisions (when there's solid grounds for that), I can usually convince them to bring up the topic upstream and get a better solution approved
<joepie91>
there just seems to be this weird and widespread perception that your job as a developer is to write the code that your superiors tell you to write
<joepie91>
it's not a strongly-held belief or anything, for most people, just... an assumption they've made, many people have made, and that never got challenged
<joepie91>
if you want to improve the quality of software overall, that is probably a good point to start :)
<andi->
Yeah. I often have such discussions about ethical points about their software... Folks realize that they might be doing something slightly the wrong way and agree but then can't change anything because they are being paid for it :/
<andi->
"I just built the bomb. I didn't make it explode."
<joepie91>
encouraging a speak-up culture
<samueldr>
ethics are hard, and generally not part of CS curriculum
<andi->
I think we talked about that before... Here in Germany they were mandatory when I studied..
<gchristensen>
omg it booted
<andi->
gchristensen: now you just have to repeat that until you fix all the typos ? :P
<gchristensen>
exactly
<joepie91>
andi-: right, so I think that's something that can be changed even without ethics courses
<joepie91>
ethics are a problem as well, but seemingly not as widespread as people feeling unable to speak up
<joepie91>
based on my own anecdata at least
<joepie91>
and people feeling unable to speak up is more a cultural thing, something for which people look to social norms, ie. influenceable in that same way...
<gchristensen>
oops, forgot a semicolon
<andi->
should have written that i haskell!! ;-)
<andi->
s/i /in /
<gchristensen>
:P
<gchristensen>
it made it past the critical section and now it is up, so I can use this box as a local cache which other servers of this type can boot much more nearby
<gchristensen>
samueldr: let's make a blockchain! github-irc-notifier -> extend that to create a log
<gchristensen>
better yet, make a git repo and host it on github
<samueldr>
gitcoin
<samueldr>
proof of work
<samueldr>
see, there, there is my PR
<samueldr>
I did work
<gchristensen>
lol
<gchristensen>
I'm looking for a proof of steak blockchain
<gchristensen>
beef burgoncoin
<colemickens>
Just another day in dystopia. Shadow banning is such a harsh thing to mess up. I don't think I'd work on implementing a shadow ban feature.
<colemickens>
oh nice riot was hours behind.
<gchristensen>
are you caught up now? :)
wildtrees has quit [Quit: Leaving]
<colemickens>
enough to see it moved elsewhere, sorta in the middle of trying to get a movie going and gettign distracted anyway
<colemickens>
but hey, I need disk space for Plex to do its thing and apparently deleting 50GB of /nix/store on ext4 over Azure IO should take 10 minutes. so here I am in IRC :)
<gchristensen>
I don't think this is shadow-banning distopia, just regular banning-users-and-deleting-their-issues-distopia
<gchristensen>
hehe
<gchristensen>
andi-: netbooting from another server in the same DC -> like 5s.
tilpner has quit [Remote host closed the connection]
tilpner has joined #nixos-chat
colemickens_ has quit [Quit: Connection closed for inactivity]
__monty__ has joined #nixos-chat
<colemickens>
Got to love the part where the accounts affected is considerably larger than the number of (successful or not) requests.
<colemickens>
Wow, and they received NSLs.
<colemickens>
This gives much to consider. One's GitHub browsing history could be revealing.
colemickens_ has joined #nixos-chat
<andi->
Yeah, I am starting to bet on https://github.com/mozilla/multi-account-containers/pull/1284 and just be not logged in most of the time and through different communication channels. Can still be correlated by just looking at a repo doesn't leave a trail..
<{^_^}>
mozilla/multi-account-containers#1284 (by samuel-crypto, 1 year ago, open): Adds proxy support as per #313
<andi->
(or not as big a trail)
waleee-cl has joined #nixos-chat
<gchristensen>
"Timed out waiting for device /dev/disk/by-id/ata-Samsung_SSD_850_EVO_M.2_250GB_S33CNX0J905800V-part3, trying to mount anyway." :(
<colemickens>
My firefox is so fingerprintable it really doesn't matter what I do unless I want to go to VMs or something ground-up like qubes
<andi->
Reminds me of the useless performance of my XPS13.. According to Dell it is just fine.. Smart also is fine.. The disk is just super slow for all kinds of operations :/
<colemickens>
That's probably overly reductive of me, I'm sure not everyone fingerprints that aggressively, and I do use containers for isolation. I've actually found that blocking all third party cookies basically causes no issues anymore.
<andi->
yeah, I also block most 3rd party resources using uMatrix.
<gchristensen>
this is a big arm box which I just installed to, and it is the only server type which didn't work correctly
<andi->
I am mostly concerned by the accidential (account, ip, repo)-tuples they might be recording
<tilpner>
andi-: Manual selection of what you block makes you fingerprint-able again :(
<andi->
tilpner: I know but you have to stop somewhere
<andi->
I actually close my browser regulary since I do not want to use the Web anymore.. It is a terrible place.
<tilpner>
I'm not saying anyone uses that for fingerprinting yet, it's probably not worth the effort
<colemickens>
I don't think they even need that full tuple though.
<andi->
Well I do not collect all the junk and do not download all the JS crap people add to their websites. If a shop doesn't load without 3rd parties I just don't buy there and close the tab.
<colemickens>
If they have logs of your account, ip, accessing a bunch of repos, then you change containers and start accessing another repo as not-your-account, it's a safe bet that's still you.
<gchristensen>
this is starting to sound a bit paranoid
<tilpner>
Oh, it is
<tilpner>
But for some people it might matter
<colemickens>
In the context of NSLs, I don't think it's out of the realm of possibility.
colemickens_ has quit [Quit: Connection closed for inactivity]
cole-h has joined #nixos-chat
infinisil has quit [Quit: Configuring ZNC, sorry for the joins/quits!]
infinisil has joined #nixos-chat
waleee-cl has quit [Quit: Connection closed for inactivity]
<evanjs>
Awe, still nothing from Antergos/web-greeter as far as I can find WRT forks. Was really hoping to see something like that on nixpkgs someday
<evanjs>
On that note, my last two distros being Gentoo and now NixOS, seeing the term “distro-hopping” (barring a select few desktop environments / greeters) just drives me crazy
<evanjs>
Like you can do _all_ of that from here without moving around all the time. What are you doingggggg
<evanjs>
Like is there really any reason people distro-hop save WMs/DEs/themes, and preinstalled apps? Setup process aside, as NixOS can make that near irrelevant...
<__monty__>
Community.
<__monty__>
And "street cred."
<samueldr>
package management paradigms
<samueldr>
different defaults in their builds
<samueldr>
reasons, not necessarily *good* reasons
<samueldr>
but yeah, any source-based distro (e.g. gentoo) or uh... quasi-source-based (nixos) distro makes all those reason pretty much go away
drakonis has joined #nixos-chat
<adisbladis>
Til: Cisco is using Nix (at least on one team)
<joepie91>
samueldr: evanjs: hmm... that's an idea... what if there were a distribution of NixOS that had different config presets replicating different distros
<joepie91>
distro.ubuntu.enable = true;
<joepie91>
that would be a fun stunt :P
<samueldr>
and maybe worthwhile, for some of their patches
<tilpner>
That might be, but I (and maybe others) associate duplicated effort with it, separate infrastructure, forums, split community, etc. Falsely, in this case
<manveru>
which is ubuntu, but with i3
<gchristensen>
because they don't have the abstractions to make it easy :)
<manveru>
exactly :)
<manveru>
one "problem" we currently have is that everyone using nixos builds their own personal snowflake distro, which is perfect for them, but kinda hard to generalize to benefit everyone without a lot of learning and effort
<gchristensen>
we do? :P
<samueldr>
I think "problem" is the right phrasing (quotes included)
<manveru>
it's why i use it in the first place :)
<manveru>
i can finally customize my system without fear of losing it all tomorrow
<samueldr>
same here
<samueldr>
maintaining the gobohide patch on top of archlinux was an horror
<samueldr>
on nixos? I forget it's there
endformationage has joined #nixos-chat
<manveru>
but i think to take this even further, we need a better way to programmatically modify nix expressions...
<gchristensen>
:o
<manveru>
atm i work around that by modifying json, but that's kinda... limited :P
<samueldr>
the idea being that a GUI to edit nixos options would make more sense then
<samueldr>
right?
<manveru>
yeah
<manveru>
and i suck at haskell too much to try hnix
<samueldr>
I didn't know it was in nix-community, that's nice
<manveru>
hmm, that could work, i've hardly done any rust, and calamares only supports c++ and python, but one can always shell out
<samueldr>
yeah, a good command-based API to modify nix expressions sure would be universally useful
<gchristensen>
it would be interesting to know how you would make meaningful changes underneath functions
<manveru>
yeah, no example for that there :|
<manveru>
but i'm mostly thinking of very simple modifications at first, we can always improve later if the proof of concept works and people like it
<drakonis>
greenspun's tenth rule here
<manveru>
just adding to systemPackages or setting a timezone would already be cool
<gchristensen>
you're already inside a function with the first 4 words :P
<samueldr>
you can probably do most good by assuming some things for nixos configurations
<samueldr>
like, look at the file, is it a function expression? check that its value is directly an attrset
<gchristensen>
I feel a gui installer doesn't need to be flexible enough for long-term use, and that getting people to a working system then saying "okay, here is how you write a nix expression and stuff you can do ... how to learn ... etc" is okay
<manveru>
for that i'm mostly trying to replace my current hacky string interpolation from a predefined configuration.nix
<manveru>
and i want to cover some common default configs, like selecting different hardware and DEs mostly, plus getting the partitioning info back into the hardware-configuration.nix
<emily>
imo a good installer would be a byproduct of a good configuration GUI
<emily>
which can probably be done not entirely terribly even with the existing metadata, it'd just need someone to actually do it
<emily>
with option.html levels of friendliness
<gchristensen>
the scope of one is an incredible amount of additional complexity
<samueldr>
it depends if we want to make a write-once configurator, or a reusable configurator
<emily>
add some highlighting/sorting for the most important options to change when installing, hook it in to the configuration generator to initialize them, and then the rest is ""just"" plugging it into partitioning
<emily>
I think giving people a clicky installer is useless if they have to immediately start writing Nix programs.
<gchristensen>
I think y'all underestimate the users
<drakonis>
y'all realize that y'all about to reinvent lisp in order to have an installer?
<gchristensen>
:D
<Church->
Damnit, beat me to it
<gchristensen>
drakonis: that is what I'm arguing against, indeed
<samueldr>
I think there is an underestimation of the cognitive weight the current nixos installation brings
<drakonis>
might as well expose the whole functionality as part of nix instead of just using it for the installer
<gchristensen>
samueldr: I agree: get them past that initial hump and then they can more comfortably learn nix
<__monty__>
gchristensen: Current users maybe but should that be the target?
<samueldr>
so I believe it's important to figure out a good way to get them started without having to write nix wholesale... but they can't be left stranded without the same configurator after the install
<gchristensen>
imo they can and that is a totally fine place to get to
<gchristensen>
it opens up who can get in to Nix and to a place they can experiment safely and easily
<samueldr>
no, since that means they need to get the configuration just right during the install, or else you need to learn more now to continue
<gchristensen>
well yes
<gchristensen>
they need to learn some stuff but they have a system they can use to get there
<samueldr>
maybe they don't
<gchristensen>
then they reboot and try again lol
<gchristensen>
which is the same thing we already do
<samueldr>
not exactly, they have a working system, but lack e.g. a web browser
<gchristensen>
time to learn!
<samueldr>
it's different from a booting system but lacking drivers to connect to the internet, which now rebooting makes sense
* gchristensen
gets on the "incremental progress" unicorn and rides away to the grocery store
<samueldr>
while true, incremental progress is good, the moment nixos gains a gui installer that is "easy to use", a bunch of well-intended reviews will appear
<samueldr>
and if you're left stranded after the install without the cozy tool, it's going to leave a mark on the never-forgetting internet
<ajs124>
The internet isn't that never forgetting.
<samueldr>
(though, not sure never-forgetting internet rings true with the dwindling quality of search results)
<gchristensen>
then let's not do it, because until we reimplement lisp, it'll be lacking
<samueldr>
lacking is not the issue, it's the stranding-ness
<samueldr>
in my opinion it's fine if we don't allow reading all existing configurations
<samueldr>
after all, it's hard
<manveru>
that parser seems to handle functions just fine btw
<samueldr>
but reading its own subset of simplified configuration to get the ball rolling
<gchristensen>
creating a pressure for all of nixos' modules to match that simplified subset
<drakonis>
shopify reimplemented lisp so they could use nix
<samueldr>
I'm not sure I follow
<drakonis>
they wrote a subset of lisp to drive nix
<gchristensen>
samueldr: you've stratified the ecosystem of things it can and can't do, and there will be pressure for the ecosystem to adapt to all be doable under whatever capabilities the tool provides
<drakonis>
it is kind of crap
<gchristensen>
anyway, I srsly do need to go to the store, so back in a bit :P
<drakonis>
also, you either implement all of lisp or you dont
<samueldr>
I'm sure I don't follow, in the context, nixos modules are not involved in my thinking. I'm thinking solely that a configurator-written configuration should be re-editable in that same configurator, or else you're sending them off the deep-end at the first boot
<samueldr>
so it's not important if, once edited, the file can only be partially re-edited by the configurator
<manveru>
exactly
<samueldr>
doing a write-once configuration is trivial enough that it's something I looked into doing, but I stopped once I realized I couldn't make it trivially editable by the GUI once generated
<samueldr>
though, that's my personal opinion, and I'd hate if my opinion alone stops someone from working on that thing
<tilpner>
An option editor for basic values should also be fairly easy, I just think it would be a pain to use
<tilpner>
We already have json files that describe the available options. For each option type (except functions, and maybe even packages) you need a different input field
<ajs124>
I would probably have the usability of regedit or the group policy editor.
<samueldr>
(my not-exactly baked idea was to transparently show the nix expressions the configuration would write in-line)
<tilpner>
So you could list all the available options, and set values for them. Submodules are tricky, no idea if that'll be as easy
<drakonis>
y'all circling around the issue
<samueldr>
drakonis: fine, we'll rewrite all nix in lisp then
<tilpner>
And you can cheat your way around generating Nix with { ... }: { config = builtins.fromJSON (builtins.readFile ...); }, if you really wanted to
<samueldr>
gosh
<drakonis>
gosh that already exists, its called guix!
<samueldr>
tilpner: I think that's what manveru has
<manveru>
jup
<samueldr>
though in my opinion that's a step in the wrong direction since you don't lead by showing an example of what to do
<samueldr>
you kind of trap the user in that weird jsony-configuration subset
<drakonis>
the user is already trapped
<srk>
manveru: hnix is already usable for rewriting nix - update-nix-fetchgit for example
<samueldr>
srk: are comments handled?
<srk>
it's a bit of a brainfuck tho
<tilpner>
samueldr: It would be trivial to generate Nix from that, once
<srk>
samueldr: I think so, that's long done
<samueldr>
they weren't last time I checked
<samueldr>
nice
<manveru>
srk: now i just need someone who speaks haskell and has motivation to work on the installer :)
<tilpner>
samueldr: But round-tripping a Nix config through a graphical editor is non-trivial, that's why JSON is in there
* srk
busy working on hnix-store :)
<samueldr>
tilpner: that's it, once, and then they lose the ability to re-open that config to further modify it graphically
<samueldr>
yeah
<samueldr>
I know it's not easy :)
<tilpner>
So they're not entirely trapped
<tilpner>
They might practically be trapped, yes
<samueldr>
well, mostly trapped, they can "eject" and lose the comforts of the editor
<samueldr>
yeah
<manveru>
rnix does round-tripping apparently, including comments
<samueldr>
that's how I got sniped jD91mZM2 in working on it
<samueldr>
how I sniped*
<manveru>
now i just have to figure out how to actually modify the AST and i'm golden :)
<tilpner>
There has to be a cutoff point
<samueldr>
(after my abject failure at trying my own nix parser)
<tilpner>
You can't roundtrip all Nix configurations
<samueldr>
tilpner: exactly
<samueldr>
you roundtrip your own, and compatible edits
<samueldr>
in an ideal world, you "blackbox" incompatible values in the attrset
<tilpner>
If there is an attrset
<tilpner>
We don't know that
<samueldr>
so they exist, editable in text, shown in the GUI as "incompatible for the user"
<samueldr>
tilpner: exactly
<manveru>
you can pull a hardware-configuration.nix and just import that generated config in your main configuration.nix anyway
<manveru>
with a nice disclaimer about modifying the file
<samueldr>
that's one of the starting point I used: you check that it's either the plain attrset or a function returning the attrset
<srk>
samueldr: hah, no, hnix won't do comments. maybe it's a feature of update-nix-fetchgit that generates spans
<manveru>
but i really think this will work anyway for users who are afraid of touching code
<samueldr>
srk: aw
<tilpner>
I'm not sure users who're afraid of touching code will have a great time with even graphically configured NixOS
<manveru>
not yet, anyway
<samueldr>
well, now we're sure they won't
<samueldr>
nudge them gently
<srk>
well you can do packagekit like app installation
<srk>
and call it a day
<tilpner>
Maybe for basic usage, but as soon as they need anything not in nixpkgs, they lost
<samueldr>
it's not necessarily afraid of touching code, but having issues wrapping their mind around it all
<tilpner>
Whereas people on other distros who are afraid of the terminal, still manage to copy-and-paste FHS-assuming build commands, for better or worse
<drakonis>
that sounds like it'd go against the whole idea, but fair enough
<tilpner>
(That may have sounded wrong, my mom is afraid of the terminal, and that's perfectly fine)
<drakonis>
y'all making some pretty deadly assumptions here though
<samueldr>
which are?
<drakonis>
first, no amount of installers will fix the usability cuts
<drakonis>
maybe if you're shooting for really simple starter configurations
<samueldr>
right, so it's entirely impossible to work on an installer while there are "usability cuts"?
<samueldr>
I figure that will happen in around 2038
<samueldr>
or the following year, 2038
<drakonis>
i'm not done yet.
claudiii has quit [Quit: Connection closed for inactivity]
<samueldr>
work can, and should be done in parallel for all of those "cuts", graphics or not
<samueldr>
uh, graphical installer or not*
<srk>
2020 is surely a year of the Linux desktop
<drakonis>
every year
<drakonis>
but if you have someone who wants to do something slightly more complex than setting up simple services and environments
<samueldr>
different people will work on different things, not every developer can fix every issue, but they can work on *new* things that are net positives while there are other issues to work on. discouraging any new work because there are issues still might just stop the developer from doing anything
<drakonis>
such as extending packages with plugins and such
<drakonis>
you'll need some sort of way to nudge them onto doing such a task
<samueldr>
sure, and that's exactly when you start learning, helped with the undestandable source your tool got you
<drakonis>
otherwise they're going to trudge along for a week or two and give up
<samueldr>
I'm still waiting for the deadly assumptions
<drakonis>
deadly wasn't the right word
<srk>
like what happens when you give bfus such powerful tool as NixOS
<drakonis>
bfus?
<samueldr>
and that is why I never spoke up about those plans, every time anyone touches up the idea of a graphical configuration tool it's shot down from all side with "but you won't be able to do that one complicated thing a small subset of users are maybe doing" and such
<srk>
drakonis: n00bs :)
<drakonis>
ah i see.
<drakonis>
write up a basic installer
<drakonis>
don't have to cover every possible permutation
<drakonis>
i'm probably being a fool.
<samueldr>
it seems the only thing that ever comes from discussing the possibility is "it won't work and you should feel bad to even think about doing it"
<emily>
22:02 <drakonis> might as well expose the whole functionality as part of nix instead of just using it for the installer
<emily>
of course, that was my point
<emily>
a config gui is a bigger step to making nixos accessible than an installer
<drakonis>
then you'd make nixlang into a lisp dialect :V
<manveru>
i'm not opposed to that :P
<emily>
like I can tell you that from having known a bunch of "non-expert-hacker" Linux users who installed NixOS fine but gave up due to the fuss of continued configuration
<emily>
Arch desensitized people to hellish install processes
<emily>
not that I don't agree it's a good thing to fix
* srk
votes for Haskell eDSL and disappears
<emily>
just I think it makes most sense as downstream work of config UIs
<drakonis>
srk: sounds fun
<drakonis>
but can all the lisp homoiconity insanity be reproduced?
<drakonis>
write the config GUI and you win
<drakonis>
end of discussion
<__monty__>
samueldr: Fwiw, I totally agree with what you're suggesting. (Just didn't want to speak up because working on that doesn't really interest me.)
<drakonis>
that'd be your installer, but comes with some kind of install script for the initial run
<drakonis>
guix has something that's a crappy installer that works decently well
<drakonis>
covers enough to get the user to xorg
<drakonis>
its like debian's installer
<samueldr>
it looks like you're ignoring that homoiconicity is not required for this to work
<drakonis>
it isnt required.
<srk>
partitioning is hard
<srk>
better re-use something existing :)
<srk>
and testing of the installer
<drakonis>
wasnt there some parser work done in the past?
<drakonis>
nixpart?
<manveru>
srk: that's why i use calamares :)
<samueldr>
and when expressions are not simple anymore, that's most likely because the user isn't working with the configuration tool anymore
* srk
worked on automated testing of Anaconda before
<srk>
opencv/tesseract for ocr
<srk>
fun stuff
<srk>
manveru: yeah, seen that few weeks ago, looks good. certainly better than doing it from scratch
__monty__ has quit [Quit: leaving]
<drakonis>
so, then what'd be the consensus, write a config gui or a installer?
<drakonis>
or write a config gui that can be used as an installer
<manveru>
i'm just doing the installer, the gui can be separate afterwards
<manveru>
and nobody really wanted to help on that, so consensus is overrated anyway :)
<drakonis>
well, i figure it might be best to just let you work on it and see how it turns out
colemickens_ has joined #nixos-chat
<manveru>
hehe
<manveru>
i'll struggle along, hope i'll have something for 20.09 or nixcon