Gentoo strategic direction re: udev

Bones McCracker · Veteran Joined: 14 Mar 2006 Posts: 1611 Location: U.S.A.

Is the way forward for Gentoo documented anywhere, with respect to udev, systemd, fork(s) of udev, sticking with OpenRC, etc.? Is there an official plan?
_________________

NeddySeagoon · Posted: Sun Jan 27, 2013 11:20 am Post subject:

BoneKracker,

eudev is being provided as an option.

There are no plans to make eudev the defualt. Indeed, Gentoo is about options.

Should it become impossible to untangle udev from systemd, then a new default may need to be selected. That will be the driver.
Until that time, its business as usual. You get to choose.

This is one of those cases when its important to keep the options open and not make a decision.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

Bones McCracker · Veteran Joined: 14 Mar 2006 Posts: 1611 Location: U.S.A.

Yes, I understand. Good plans often involve developing information, reducing uncertainty, and then making decisions.

So, does "untangle" mean obtaining an unambiguous commitment, from whoever governs the architecture of systemd and udev, to modularity, enabling distributions using init/rc systems other than systemd to make architecturally clean, unencumbered use of udev for the foreseeable future?

I get the sense, from one statement he made, that Mr. Poettering's vision is a world where everybody eventually migrates to systemd. After people raised some fuss, he has spent a lot of time making carefully-worded statements indicating udev remains effectively seperable and things will still work as is "for a long time". But, frankly, I don't see any commitment to a permanently seperable architecture. I'd rather know now than be lied to.

Does openly talking about this create problems?
_________________

mv · Watchman Joined: 20 Apr 2005 Posts: 6780

aCOSwt

mv · Watchman Joined: 20 Apr 2005 Posts: 6780

Bones McCracker · Veteran Joined: 14 Mar 2006 Posts: 1611 Location: U.S.A.

Okay, let's all hold on here.

The last thing I want to do is start yet another ranting session. I'm sure that what Mr. Poettering is doing makes sense from his perspective, although it may be the desire of others that udev be cleanly separable. Without making any judgements, I for one do not believe it is his intent they remain cleanly separable indefinitely.

I think all discussions people are having about this will be more productive if they don't get too personal in a critical sense.

I don't even necessarily want to know what Gentoo's plan is; mostly I'd just like some reassurance that there is one. Beyond that, it might be nice to know what it is (but I can understand there may be reasons why it not to be public knowledge).

But, really, it's enough for me to know what NeddySeagon said, which is essentially that we're keeping our options open at this point, and I do not want to stir up even more discord.
_________________

SamuliSuominen · Posted: Sun Jan 27, 2013 3:59 pm Post subject:

Bones McCracker · Veteran Joined: 14 Mar 2006 Posts: 1611 Location: U.S.A.

So you think it will remain reasonably separable and current. What about this:

NeddySeagoon · Posted: Sun Jan 27, 2013 7:57 pm Post subject:

BoneKracker,

While udev can be used without systemd there is no need to plan to do anything. If we wake up one day and find that udev is no longer separable from systemd, they Gentoo needs to make a choice. Until then, its business as usual.

There are several choices now, depending on your needs
systemd (yes, it really is a choice)
eudev
mdev
static /dev
Anything else in the udev virtual.
There is also the short term, use the last useful version of udev. That buys us a month or two to plan.

Personally, I hope the good idea fairy pays the systemd devs another visit and they get to understand that 'vertical integration' is a terrible idea.
I've been there, done that and seen a project scrapped because with 'vertical integration' the code was unmaintainable. It couldn't even be debugged well enough to make a first delivery. On the basis of that experience I have systemd hard masked.

I hope that a choosing among udev alternatives will not be required any time soon, if ever, but like you, being prepared is always a good idea.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

666threesixes666 · Posted: Sun Jan 27, 2013 8:02 pm Post subject:

im using systemd, only loss so far is apache... vsftpd unit file i drew up from scratch (and permit to cross over from gentoo-wiki to wiki.gentoo.org) works well. can a dev plz start a unit file hoard on the wiki for me to add to.....

_______0 · Guru Joined: 15 Oct 2012 Posts: 521

as just a user I think systemd main fault is to be a binary controlling ALL. That remains the soles point of failure.

Latest systemd's/Potter 'propaganda' effort is his 'debunking 30 myths' on his blog and all.

True, each of those 30 points do solve something and are amazing solutions.

Yet, cramming ALL those pieces into a single binary doesn't seem to be a very wise idea.

The init scripts are pieces of script that manually bring up services, if a service screws up doesn't affect the rest. Looks like having scripts plus openrc (more manual) adds two steps to ensure things work. Whereas with systemd's unification zeal has the potential to do to many things for itself (which seems to be the road is taking). 'ooooh aahh systemd get's rid of a,b, c, d, f ... tasks into a single binary!!!'

Besides, do you really need a binary to this?

modprobe foo

lololol

In anycase it remains to bee seen how systemd fares, only time will tell.

The controversy lies in the difference of paradigm I think, binary versus scrips. On phoronix one commenter call's systemd m$$$ registry implementation on linux.

I have to re-read his 'debunking 30 myths'.

mv · Watchman Joined: 20 Apr 2005 Posts: 6780

Element Dave · Tux's lil' helper Joined: 10 Nov 2006 Posts: 82

Unfortunately my message text got "lost" and I can't retype everything at the moment, but briefly I wanted to comment that I was surprised by the recent decision to stabilize udev-197 and the abrupt proposal to migrate to the new systemd/udevd network interface naming scheme. The new network interface naming scheme was only proposed just the other day for Fedora 19 on the Fedora-dev list, and by no means seemed certain even there. And this is Gentoo.

What exactly was the problem with udev-17x that -197 solves? The new udev seems much buggier, while the older one was working fine even with the most recent kernels. Hmm.. Lots of executables now link to libsystemd-daemon.so.0 too.

More later...

SamuliSuominen · Posted: Mon Jan 28, 2013 8:49 am Post subject:

yngwin · Posted: Mon Jan 28, 2013 9:47 am Post subject:

Bones McCracker · Veteran Joined: 14 Mar 2006 Posts: 1611 Location: U.S.A.

Thank you for these valuable insights. That pretty much fills in the gaps for me.

Also, I don't mean to be critical of the lack of a "plan" per se. Loosely governed anarchy has its benefits along with its drawbacks (and I mean "anarchy" there without any of the usual negative connotations; I mean "highly decentralized governance).

Maybe I should monitor the dev mailing list if I'm interested in such things, rather than asking whiny questions in here and setting off more rants. Thank you both, and Neddy, for taking the time to answer.
_________________

slackline · Posted: Mon Jan 28, 2013 11:03 am Post subject:

There's some discussion on udev over on Reddit.
_________________
"Science is what we understand well enough to explain to a computer. Art is everything else we do." - Donald Knuth

Bones McCracker · Veteran Joined: 14 Mar 2006 Posts: 1611 Location: U.S.A.

mv · Watchman Joined: 20 Apr 2005 Posts: 6780

Bones McCracker · Veteran Joined: 14 Mar 2006 Posts: 1611 Location: U.S.A.

That individual apparently needs to read more of what they have said. I think they've pretty clearly stated their intentions. It's possible that they've got so much grief about it they may feel pressured into maintaining explicit modularity, and maybe they need help to be able to do so, but from what I see, the intent is clear: use systemd or you're on your own as far as getting udev to function as a separate product.

That individual's "could have done that already" logic makes no sense and fails to recognize the evolutionary nature of the changes being made (which appear to me to be intended to gradually compel everyone to migrate to systemd). Now, maybe doing that would not be a bad thing necessarily, but people need to be honest about things.

How much tangible involvement in the eudev project has emerged from non-Gentoo folks?
_________________

Hypnos · Posted: Mon Jan 28, 2013 3:39 pm Post subject:

Reading ajax's take on choice, and those of his Red Hat compatriots in the follow-ups, it is clear that Red Hat and Gentoo have divergent goals/principles. Gentoo sees Linux as an ecosystem, Red Hat sees it as a platform for *their* OS.
_________________
Personal overlay | Simple backup scheme

luismw · Tux's lil' helper Joined: 04 Jan 2010 Posts: 91

After reading yngwin's explanation of how Gentoo works, I still have one doubt. I can understand that developers can work freely on whatever they like and so on, however, someone has to make a decision on things such as default package choices or, for instance, the init system. Who decides this? The council? A meeting of developers? Is it done according to a plan or just on a case by case basis?

Just out of curiosity.

John R. Graham · Posted: Mon Jan 28, 2013 5:37 pm Post subject:

As described above, Gentoo Developers work on what interests them. A lot of those foundational choices are decided by members of the Gentoo Base System Project. Whenever a major change is proposed, there's lots of lively discussion on the gentoo-dev mailing list.

- John
_________________
I can confirm that I have received between 0 and 499 National Security Letters.

hcaulfield57 · Tux's lil' helper Joined: 13 Mar 2012 Posts: 148