DiffieHellman | >Let's keep eudev maintained >github.com Could ya'll please consider not using a proprietary git hosting service? | 03:40 |
---|---|---|
DiffieHellman | It's a bad look for a project that escapes the clutches of systemd to throw itself into the clutches of githuw. | 03:41 |
UsL | DiffieHellman: https://p.pastly.xyz/file/zn75n2/4n4lgn/plain | 03:55 |
DiffieHellman | UsL: Gitlab isn't a good choice either, it's only slightly better. | 04:01 |
DiffieHellman | I would be happy to host gitea for you if I could guarantee uptime - sadly I can't yet. Best way is just to use cgit and have a mailing list :D | 04:03 |
UsL | not sure that would be the easiest path for the project. I guess they settled for github for the reasons in the paste. | 04:05 |
UsL | I'm just glad eudev will continue now and that devs have a platform. | 04:06 |
DiffieHellman | If you've settled on github, at least accept patches via git send-email I reckon. | 04:07 |
UsL | they probably will cater to most needs out there. | 04:09 |
plasma41 | Was SourceHut considered? I is fully functional without javascript and email-based workflows as supported as first-class. | 04:10 |
plasma41 | s/I/It/ | 04:10 |
DiffieHellman | Well, I want to send a patch in to fix the readme, but I can't. | 04:11 |
DiffieHellman | I.e `IRC: Freenode/#gentoo-udev` - that channel is on librechat now | 04:12 |
UsL | they will soon forward here if I recall correctly. Like devuan-eudev got +F here. | 04:13 |
UsL | I think it was discussed earler today/yesterday | 04:13 |
UsL | plasma41: sounds good. No one spoke of it I think. Perhaps bring it up with them if it's a better fit | 04:15 |
plasma41 | UsL: Who is "them" in this case? Are you referring to anyone who isn't in this channel? | 04:16 |
UsL | from the paste link I pasted some rows up | 04:16 |
plasma41 | Ah, in that case: | 04:19 |
plasma41 | ping Ariadne and bb|hcb ^^^ | 04:19 |
UsL | and Arsen. | 04:20 |
UsL | he already pulled one year of commit stuff to a thingie. | 04:21 |
UsL | anyway, checkout the logbot link. No need to repeat | 04:21 |
* plasma41 reads through the log | 04:24 | |
golinux | DiffieHellman: librechat > liberachat | 04:40 |
DiffieHellman | Damn, I slipped up, unacceptable. | 04:44 |
golinux | It happens. :) | 04:44 |
golinux | No biggie except to someone who might try to c/p | 04:45 |
Ariadne | we are not interested in bikeshedding the git forge. thanks | 06:26 |
DiffieHellman | You're only interested in accepting commits from those willing to surrender their freedom to microsoft? Okay then. | 06:28 |
Ariadne | literally not here to deal with concern trolls | 06:47 |
UsL | nice attitude. Perhaps more becoming in a mid level ceo at - insert big bully corp here - than a extremely small project with two devs. If he wanted to post patches his way by mail that makes him a troll? You should perhaps also speak for yourself and not use "we" in your sentences. | 08:21 |
Arsen | DiffieHellman: the clutches of a git host matter not | 10:10 |
Arsen | it's a distributed scm | 10:10 |
Arsen | ah they left | 10:11 |
joerg | while I'm also not happy with gitblub, I admit it's more easy to switch that some sourcecode once they go rogue (we seen a similar switch lately, leenode anybody?) | 10:19 |
joerg | than some* | 10:19 |
skarnet | please let's not have yet another iteration of this conversation | 10:25 |
Arsen | yeah | 10:35 |
Arsen | fwiw we could accommodate for people who boycott GitHub because of Microsoft by having a mailing list somewhere | 10:52 |
Arsen | pull requests and emails can both be sent in a standard git manner to such a list | 10:53 |
Arsen | and patches, I mean | 10:59 |
* rkta likes having the option to send patches via mail | 11:00 | |
Arsen | I think at least two of us here are accommodated to that workflow, dunno about the rest | 11:00 |
Arsen | skarnet: what do you think? iirc you've worked like that for quite a long time | 11:00 |
skarnet | Arsen: it's a problem that Alpine has struggled with for a long time, I'm certainly not the best person to talk about it, but from what I saw, maintaining several workflows was a lot of effort, mainly because ml-to-PR software was not up to speed, etc. etc. | 11:10 |
skarnet | my general opinion in these matters is that people spend waaaaaaaay too much time thinking about meta stuff and talking about what would theoretically be best, and when (if) they eventually make a decision, implementing/integrating/managing the tooling that would possibly be useful in the theoretical cases etc. | 11:12 |
skarnet | and in the end much more effort is spent on that than on actual stuff that needs doing | 11:13 |
skarnet | and it's a pattern that infuriates me | 11:13 |
skarnet | experience has changed my opinion of workflow/hosting/freedom... purists | 11:14 |
skarnet | I was one of them | 11:14 |
skarnet | nowadays I think people who insist on that kind of purity simply lack maturity | 11:14 |
skarnet | the concern about using GitHub is real, but needs to be quantified | 11:15 |
skarnet | how likely is it that MS will pull a fast one and deprive github users of their freedom, whatever that means | 11:15 |
skarnet | what would that mean for eudev | 11:16 |
skarnet | what would be the cost | 11:16 |
Arsen | well let's answer that one by one | 11:16 |
skarnet | what is the cost of an alternative | 11:16 |
Arsen | - MS will never pull anything on GitHub because it's their enterprise platform | 11:16 |
skarnet | how much effort is needed to set up alternative workflows | 11:16 |
Arsen | - it would mean nothing, git is the easiest thing to migrate, hell, I'll set up a mirror for our eudev repo this week just in case, that we can recover from whenever, as I do for many other repos | 11:17 |
Arsen | - I don't think the cost can exceed one workday in manhours | 11:17 |
Arsen | no matter what happens | 11:17 |
Arsen | - not much, actually | 11:17 |
Arsen | as for ml-to-PR, I don't think such a thing would happen | 11:17 |
skarnet | I'm not looking for answers, I'm just listing the questions that I think people should ask themselves whenever they complain about lack of purity | 11:18 |
Arsen | ah, my bad | 11:18 |
skarnet | just exposing my thought process on these kinds of things | 11:18 |
Arsen | but, back on that: I assume everyone here has good mail netiquette (email accounts, clients set up for plain-text mail, etc), so involving a review process on that alternate channel wouldn't be an insane proposition | 11:19 |
Arsen | nor too difficult, really | 11:20 |
Arsen | automatically converting them to PRs doesn't sound like a good idea to me, but I could figure that out | 11:20 |
Arsen | it would mess up committer information in the commits and would open up a channel for abuse | 11:21 |
Arsen | not to mention it'd probably be relatively fragile | 11:21 |
skarnet | and unless there's obvious reason for immediate or short-term concern, or fundamental/ethical reasons, my general approach is "stop jerking off to purity and getting nothing done, make a choice, do some real work, if it appears that the choice was bad, then spend effort changing it, but don't spend all your time overthinking it" | 11:21 |
skarnet | fundamental reasons can be e.g. dogfooding: I run skarnet.org on my own software. I expect a forge to self-host. etc. etc. | 11:22 |
skarnet | but those are special cases | 11:22 |
skarnet | I'm interested in the content, not on the metadata | 11:23 |
skarnet | not in* | 11:23 |
skarnet | and in tech people spend their LIVES talking about metadata | 11:23 |
skarnet | that was a very long answer to eventually say "I really don't care, and you shouldn't care either" | 11:24 |
skarnet | do whatever you want, I'll adapt | 11:25 |
Ariadne | UsL: thank you for your feedback, let me summarize my response: don't care | 11:28 |
joerg | :-) I wasn't aware there was so much fuel to burn on this topic. I agree those considerations are pretty much proven by experience (also my own, regarding a lot of tools). Probably a good thing it got diskussed, evaluated and the result archived now ;-) Back on track with main topic for me | 11:29 |
Ariadne | skarnet: the choice of forge was not my decision, github was chosen as lowest common denominator. it was what everyone could agree to hate mutually. | 11:29 |
Arsen | it ultimately really doesn't matter | 11:31 |
Arsen | it's beyond simple to pick and switch | 11:32 |
Ariadne | exactly | 11:32 |
skarnet | Ariadne: that's missing a fundamental government principle: in order to have the least possible amount of discontent, you should always oppress the same people! :P | 11:36 |
UsL | Ariadne: what you care about is evident. My point was that perhaps inclusion is better than sitting on a high horse dictating how people are not welcomed to contribute to this project | 11:40 |
Ariadne | ok | 12:47 |
bb|hcb | UsL: crying for how an accident may happen to the unborn baby is pointless | 14:19 |
bb|hcb | If there is a person who has to avoid github (whatever their reasons), we will find a way to solve that | 14:19 |
bb|hcb | Until there is such a problem to solve, please do not put any effort in it | 14:20 |
bb|hcb | And to (hopefully) finish that topic: There are 104 forks, 26 open issues and 1 PR on github already. Shall we cut off all those people because we dislike github (not a question) | 14:24 |
joerg | [17 Sep 2021 06:28:29] <DiffieHellman> You're only interested in accepting commits from those willing to surrender their freedom to microsoft? Okay then. | 15:28 |
joerg | [17 Sep 2021 06:28:29] <-- DiffieHellman (~Username@user/curve25519chacha) has left this channel. | 15:28 |
joerg | I also think a more inclusive approach to complaining users is advisable - at least in message they are given by the channel | 15:30 |
zhadum | How do you surrender your freedom to Microsoft by using github? The mind boggles | 15:31 |
joerg | questionable statement, yeah. However no reason to instantly leave the channel | 15:32 |
skarnet | I question the size of { potential contributors of value to eudev } ∩ { grandstanding purists } | 15:33 |
joerg | please keep in mind we're talking about a recent actual context: http://reisenweber.net/irclogs/freenode/_eudev/_eudev.2021-09-17.log.html#t2021-09-17T06:28:29 | 15:35 |
joerg | I'm not really interested in gitblub, I am dedicated to the mood in this channel | 15:37 |
joerg | maybe we could agree to give users like DiffieHellman a "don't worry, we'll find a way you feel comfortable with" reply? | 15:39 |
skarnet | the point is that I suspect most of these users would not actually contribute to eudev in any meaningful capacity and are safe to ignore | 15:43 |
skarnet | there are ways to express that you are interested in participating to a project but uncomfortable with the current tooling | 15:44 |
skarnet | and this is not what happened | 15:45 |
joerg | I guess you're right, as their introduction here was >><DiffieHellman> >Let's keep eudev maintained >github.com Could ya'll please consider not using a proprietary git hosting service? << I don't know what if any been the history before that point in other channels with this user. | 15:47 |
zhadum | How come these people stick with Freenode and then complain about gitbub? How come this channel's logs even available from Freenode? Is there a relay? | 15:47 |
joerg | sorry? | 15:49 |
zhadum | I am in a LiberaChat channel, how come the log URL is a Freenode channel log? | 15:49 |
joerg | it's not. I guess you mix my logbot's local network names which for hysteric raisins are somewhat misleading with what's actually getting done and where | 15:50 |
zhadum | I was k-lined on Freenode just for what I represent and voicing my concern about Lee's takeover and his shadowy admin hire process. So it is comforting to hear that it's just a bot misconfiguration, not some leakage from Freenode ;-) | 15:51 |
joerg | so welcome to the club, I moved over like a dozen channels from leenode to here | 15:52 |
zhadum | Same | 15:52 |
zhadum | But, I promised myself a break from online communities just two weeks ago and here I am, discussing again. So I'll try to keep my mouth shut now | 15:53 |
joerg | I guess eventually I got to actually rename that stuff. Does anybody know of a tool like sed that *also* acts on filenames? | 15:54 |
Ariadne | perl has rename(1) | 16:12 |
joerg | ta, alas my perl skills are really poor | 16:13 |
Ariadne | zhadum: i feel like the crown prince of the joseon empire and his IRC network are unrelated to eudev | 16:13 |
joerg | guess I'll write a bash script. like always | 16:13 |
joerg | Ariadne: it was about chanlog in /topic | 16:14 |
Ariadne | fair | 16:18 |
skarnet | joerg: find -exec mv ? | 16:20 |
skarnet | (in addition to the sed) | 16:20 |
joerg | sth like that, yes | 16:20 |
joerg | will be like 4 basg lines in the end | 16:21 |
joerg | bash even. Then lighttpd rewrite rules on top of this for legacy bookmarks :-/ | 16:21 |
Arsen | joerg: moreutils has a neat program called "vidir" | 16:51 |
joerg | Arsen: awesome :-) | 16:53 |
joerg | I was already musing how to code something _very_ similar :-D https://linux.die.net/man/1/vidir | 16:58 |
rkta | re: sed for file names: zsh has zmv | 17:17 |
Arsen | why is that in zsh? | 17:18 |
rkta | I dunno, but zsh has it :) | 17:19 |
joerg | find --name '*abc*' -exec mv {} $(echo {}|sed 's/abc/xyz/g')\; ; ? | 18:01 |
joerg | plus sed -i 's/abc/xyz/g' $(find -type f) | 18:02 |
joerg | multiple times, in carefully sorted order of sequence, for s/irc.freenode.net/chat.libera.dev/ s/[Ff]ree[Nn]ode/Libera/ and whatnot else I forgot | 18:05 |
joerg | TLD .dev ;-P me needs more coffee | 18:28 |
Arsen | I don't think that's how find works, I think | 19:03 |
Guest12 | It's great to hear about the recent transition and thank you to everyone helping to maintain eudev and keeping all of us Slack(ers) systemd free as we like it to be. | 19:04 |
bb|hcb | #gentoo-udev will continue to be active but focusing on systemd separated udev; in this case I see no point in leaving notes or making redirects | 20:30 |
bb|hcb | Arsen: https://git.aarsen.me/systemd-udevonly.git/ gives warning: remote HEAD refers to nonexistent ref, unable to checkout. | 20:37 |
bb|hcb | Do I get something wrong? | 20:37 |
Arsen | hm | 20:38 |
Arsen | that's strange | 20:38 |
Arsen | are you cloning main? | 20:38 |
Arsen | I probably didn't adjust HEAD to main | 20:38 |
Arsen | since git defaults to master | 20:39 |
bb|hcb | I tried than on a Debian/sid vm; no changes in git conf besides signed commits, name, etc. | 20:40 |
Arsen | I meant on my machine | 20:46 |
Arsen | bb|hcb: git clone -b main https://git.aarsen.me/systemd-udevonly.git/ was fine for me | 21:10 |
Arsen | though that uses the dumb http transport | 21:11 |
Arsen | I should probably do something about that | 21:11 |
bb|hcb | I cloned it, but I am still wondering how to use that against eudev repo | 21:35 |
* bb|hcb feeling dumb | 21:36 | |
Arsen | don't feel dumb | 21:59 |
Arsen | I replaced the eudev repo with that | 21:59 |
Arsen | as I do believe a viable strategy is making a new fork off of systemd right now | 22:00 |
* sam_ nods | 22:04 | |
bb|hcb | Yes, there are two approaches - 1. generate a patch set agains current eudev code and 2. do a new fork (what you propose above) | 22:05 |
Arsen | I mean, technically, one could git merge -X theirs new-fork-branch, or something along those lines, but I'm not sure how worth it it is | 22:06 |
bb|hcb | Both have pros and cons, I can't say which is better | 22:06 |
Arsen | though, the theirs merge strategy probably won't discard other paths in tree that are no longer relevant | 22:06 |
Arsen | bb|hcb: I corrected HEAD on my server to point to main, sorry about that oversight on my part | 22:08 |
Arsen | frankly, I probably noticed it was wrong and just ignored it when I was setting that up | 22:08 |
bb|hcb | no worries - that is a detail :) | 22:09 |
bb|hcb | OK, lets brainstorm path 1; what do you propose to do next? how can others help with that? | 22:09 |
Arsen | well, while we were discussing this, a friend of mine made an interesting advancement: https://github.com/freeinit/libfreeinit | 22:10 |
Arsen | now, that repository as it stands now is empty, but I'm told that it's meant to be a portable libsystemd reimplementation | 22:11 |
Arsen | if (when) that comes to fruition, we could just depend on that in eudev and make our job entirely just keeping up with upstreams (sane) decisions, and correcting any less sane ones and unportabilities | 22:11 |
Arsen | and, of course, contributing to libfreeinit, as we would likely become a large consumer of it | 22:12 |
Arsen | "reimplementation" is the wrong word, it's probably to libsystemd what eudev is to systemd-udev | 22:12 |
Arsen | that was quick | 22:13 |
bb|hcb | ... a decoupling fork IOW | 22:13 |
Arsen | netbsduser here made that repository | 22:14 |
netbsduser | i've got all my things open, easy to context switch :) | 22:14 |
bb|hcb | Hello netbsduser and welcome! | 22:14 |
Arsen | now, what I propose as a short and long term strategy is: keep eudev alive as is, fix any bugs incoming from gentoo, alpine, devuan, and in the future, rebase onto systemd, with libfreeinit as the main libsystemd implementation | 22:15 |
Arsen | that rebase step would involve reorganizing the code in a usable form, like I've shown before via that filter, removing any glibcisms, and anything that might not yet work with libfreeinit, though our goal should be to correct that in libfreeinit, to an extent appropriate | 22:16 |
Arsen | this boils our work down to what I said before, merging in useful changes and removing problematic ones, and ensuring portability | 22:17 |
Arsen | (plus backcontributing to an independent libsystemd implementation for a greater good) | 22:17 |
Arsen | the first example of how such a thing comes into play is having libfreeinit adjust sd-notify to work with skarnet's s6-rc (excuse me if I misattributed the component - my memory is quite poor), for instance, allowing for a richer state of various services and dependencies between them | 22:18 |
bb|hcb | Basically short term plan is to check existing issues and PRs, merge distro specific patches in the current codebase? | 22:20 |
bb|hcb | And long term plan is to do a eudev-5 branch like the 5y old eudev-4 one? | 22:20 |
Arsen | yup | 22:20 |
Arsen | not all distro specific patches, though | 22:21 |
bb|hcb | Did I say all :P | 22:21 |
Arsen | for instance, I'd not be one to disable predictable interface naming, it's quite useful in many environments, and there's no need to diverge such an easy to change default in our downstream rather than distro specific downstreams | 22:21 |
bb|hcb | About interface naming, I believe that we shall make the selection to be configurable. Preferrably runtime, although its acceptable build time... | 22:23 |
Arsen | that works for me too | 22:23 |
bb|hcb | <rant>I see nothing predictable in that naming</rant> but I see no reason people to like it and use it | 22:25 |
Arsen | put 2 nics in the same pc and ens1 is always the first pci express hotplug slot | 22:27 |
Arsen | put 2 nics in the same pc, and you could very well see eth0 and eth1 swap places | 22:27 |
Arsen | also, ``ln -s /dev/null /etc/systemd/network/99-default.link'' has the same effect with zero patches, though I haven't had a chance to try that on eudev | 22:28 |
netbsduser | Arsen: that case wrt sd-daemon's sd_notify functionality is exactly one of the things i'm very keen on - i've seen too many daemons which play nicely with systemd but don't with other service supervisors, insisting on e.g. double-forking and thus breaking the likes of s6 (not sure if s6 has that awful 'fghack' thing that daemontools had which tries to keep track of double-forkers) | 22:28 |
bb|hcb | Somewhere blueness mentioned that the main thing in which current codebase is visibly behind is hwdb parsing; maybe its a good idea to look into that first? | 22:29 |
bb|hcb | Seen that, been there... But some bios renumber pci busses too, then ens1/ens2 would also swap (that was 10y back some 5000 intel mobo) | 22:31 |
bb|hcb | Arsen: can't check the 'ln -s ...' thing - eudev is already pathched on my pc | 22:36 |
bb|hcb | netbsduser: What do you think about socket activation? | 22:37 |
netbsduser | bb|hcb: i have a lot of respect for dave zarzycki (the original engineer responsible for apple's launchd) and i think that he had a great idea in using it to eliminate the need for explicit dependency directives. i think its benefits for systems which don't have that pure model, however, are more limited. i've seen sparse evidence that it's really that much of a speed gain, which is what the systemd people seem to prize as its greatest merit | 22:40 |
Arsen | yeah, you can only get as good as the fw/hw permit, but why not bother going to that limit for a few broken fws which are, at worst, just as good | 22:40 |
Arsen | I probably worded that poorly but I hope my point comes across since I have to leave the house again | 22:41 |
netbsduser | some of the remaining merits rely on the extended definition of 'socket activation' that systemd uses. sending your FDs to a trustee while your daemon restarts and then retrieving them to maintain uninterrupted service is a nice thing | 22:42 |
bb|hcb | I didn't know the idea comes from launchd. And since modern OSs have SCM_RIGHTS it is not black magic to implement. But the library interfaces and decoupled init system messaging are a problem | 22:45 |
netbsduser | strictly speaking it's inetd that pioneered it; launchd just found a clever use of it, by which launchd never had to deal with a significant chunk of the problem of dependencies | 22:47 |
netbsduser | i still need to investigate s6 more fully, then i can figure out whether it's going to be feasible to implement any useful abstraction of sd_notify that can work with s6 too. certainly some of the features of sd_notify liek the status messages won't work. and whether the sd_notify fd-holding functionality can be adapted to fit with s6-fdholder is another question | 22:50 |
bb|hcb | early unix tools used stdin/stdout and inetd made them networked; instead of improving inetd daemons started to implement sophisticaed socket management models... in some way launchd went back to the roots | 22:51 |
Arsen | this sophistication can come at an advantage, though | 22:52 |
Arsen | sharing state was easier, and handling many thousands of connections became less relevant on the CPUs ability to be concurrent | 22:52 |
netbsduser | if issues of fundamental incompatibility arrive, the logical step is to try and implement some new interface in libfreeinit, a frontend to either sd_notify style, s6 style, or whatever, and push for daemon-writers to support it | 22:52 |
netbsduser | in fact i see setting standards as an imperative. we musn't facilitate our own irrelevance by encouraging people to depend ever more intimately on the subtleties of systemd and its interfaces | 22:56 |
netbsduser | i am very enthusiastic about libseat, for example, which is a library providing the functionality sought by wayland compositors and desktop environments through an interface sufficiently well-thought-out to be able to deal with either seatd or (e)logind | 22:57 |
bb|hcb | Agree. But that is a tough to keep balance between staying practicle and sane | 23:13 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!