Fedora defers systemd to F15 [LWN.net]

Fedora defers systemd to F15

Posted Sep 15, 2010 17:18 UTC (Wed) by drag (guest, #31333) [Link] (9 responses)

Pretty much. If you don't want to deal with changes then use CentOS or Redhat.

That's the point of them. The last thing we need is another CentOS.

Fedora defers systemd to F15

Posted Sep 15, 2010 19:22 UTC (Wed) by marcH (subscriber, #57642) [Link] (8 responses)

> If you don't want to deal with changes then use CentOS or Redhat.

There is an enormous distance between Fedora and CentOS, much bigger than what was discussed here.

Fedora defers systemd to F15

Posted Sep 16, 2010 2:02 UTC (Thu) by tialaramex (subscriber, #21167) [Link] (4 responses)

Indeed, and are there also an enormous number of additional volunteers to develop something more to your taste?

Fedora defers systemd to F15

Posted Sep 16, 2010 6:36 UTC (Thu) by marcH (subscriber, #57642) [Link] (3 responses)

What you still do not get:
- the vast majority of Fedora users (not just me) did NOT ask for systemd, not even for improvements to the current system, whatever ugly it was.
- it takes ZERO volunteer not to develop systemd. How demanding for a taste?

I am totally convinced Lennart is a fantastic developer. Sorry, but "Ooooh! aaaahhh!" does not give him an automatic right to push his software fast forward.

Fedora defers systemd to F15

Posted Sep 16, 2010 10:07 UTC (Thu) by bojan (subscriber, #14302) [Link] (1 responses)

> - the vast majority of Fedora users (not just me) did NOT ask for systemd, not even for improvements to the current system, whatever ugly it was.

The vast majority of Fedora users have no idea what an init system does and why it is important to have a truly dynamic one. Please go and view that Apple presentation on launchd. Apple is not in the business of pleasing geeks for geeks' sake. They switched to launchd because it is ultimately better for users and developers.

Fedora defers systemd to F15

Posted Sep 27, 2010 16:14 UTC (Mon) by topher (guest, #2223) [Link]

True, but Fedora also already switched to upstart, which provides a similar framework for that dynamic and "cool" system startup.

Personally, I'm not yet convinced that systemd is actually an improvement over upstart, so forcing it through as the new default because it "works pretty well, and is mostly reliable" doesn't make a lot of sense. I'm still waiting for a better explanation of why I should want it.

(Yes, I've read the docs on it. Some of it is looks kinda cool. . . and like a big improvement over the old SysV init. And like a marginal improvement, at best, over upstart, which is more established and significantly better tested. And, as someone who has to use multiple distributions, consistency counts for far more than any of the minor potential improvements. I don't want to have to deal with yet another init system on another distribution.)

Fedora defers systemd to F15

Posted Sep 16, 2010 20:53 UTC (Thu) by tialaramex (subscriber, #21167) [Link]

Fedora 14 will introduce numerous package updates, new features, and bug fixes and not a single one of them will have achieved your arbitrary threshold of "the vast majority of Fedora users" asking for it.

Raymond Chen has written a number of thought experiments which are valuable for those who want to understand why their idea is stupid before they propose it. The relevant one here is not "Features do not exist by default" (as you seem to think) but "What if two people did this?"

If you are entitled to a systemd-less Fedora just because not making systemd would have been zero effort, obviously all other Fedora users are also entitled to new versions which lack whatever features they don't happen to want.

But all these extra versions of Fedora require considerable integration and maintenance effort. The project will inevitably collapse if this approach is taken. On the other hand, Fedora is pointless if it each new versions just consists of a few uncontroversial bug fixes against the previous one.

So Fedora has a process to decide which new features land. In those parts of this discussion which aren't taken up with you ironically asking other people to "get a life", it is clear that the process failed, badly.

'distance' is not the correct term here.

Posted Sep 16, 2010 11:23 UTC (Thu) by drag (guest, #31333) [Link] (2 responses)

> There is an enormous distance between Fedora and CentOS, much bigger than what was discussed here.

Not ultimately. There are a hundred thousand new ways that Fedora could introduce new functionality that would break stuff for users.

Fedora is one of the last 'happy places' for individual developers and companies that really want to get moving with new technology and have a audience that is willing to test it.

Break something in Mesa, Dbus, GTK, PulseAudio, PHP, Apache, or Nautilus or Packagekit or any of that stuff and you'll piss end users off just as hard as if you broke something in their init scripts. The OS purpose in life is nothing more then to serve the purposes of developers and their applications. End users only care about their applications... whether it's a web forum software or Gimp that is what matters most. If a PHP lib breaks or a init script gets launched out of order the effect is the same for a person that wants to host a website.

There really is no logical dividing point of saying "Well this change might piss of 70% of people so it won't go in, but this change only pisses off 30% of people then it's ok'.

Were is the line you draw?

Logically the line to draw is:
Either you accept changes that you think will probably work or you do not and only accept proven changes!

If somebody has put in the work to design a superior PID 0 from scratch, meets the obligations and goals set forth before him and then still gets all his potential end users denied from him and users are denied the better software... how does that make sense? How is that behavior on the part of the denier justifiable?

It is not, really.

'distance' is not the correct term here.

Posted Sep 16, 2010 11:35 UTC (Thu) by drag (guest, #31333) [Link]

Oh, and before somebody says something silly like 'init breaking is more serious then gtk breaking!'

Well, it's not. As I explained above the kernel breaking, the drivers sucking, the init sucking, the sound server sucking, the application lib sucking, etc etc. The effect is always going to be the same to the end user:

My application is broke, the OS is shit. The user is DOS'd from their application due to a bug.

And the approach for a end user dealing with the issue the issue is always going to be roughly the same:

1. Hack around the problem.
2. Bitch about it.
3. File a bug report
4. Wait a couple weeks, install a update that fixes the problem.
5. Figure out how to roll back your hacks
6. Bitch some more about it.
7. Gradually forget that it ever happened in the first place.

I know there is different levels of pissed off-ness users will get and I know that there are big differences in the difficulty of working around a broken OS component.. but, frankly, I know I'd have a much easier time dealing with a init script then I would with php_mod or a broken Mesa driver.

So how you approach things is a question of goals. Are changes done with reasonable efforts to assure correctness is acceptable, or are then not and the software must be proven elsewhere first?

That is the difference between CentOS/Rehdat and Fedora.

'distance' is not the correct term here.

Posted Sep 16, 2010 14:06 UTC (Thu) by marcH (subscriber, #57642) [Link]

Your elaborated and interesting analysis is yet missing one of the most important parameters: time (or speed if you prefer). In this case, six months.

The number of people in this forum that seem to believe a gas pedal has only two positions is truly amazing. Damn, I just made a car analogy. (As an anecdote, I have seen a few people actually using their gas pedal like this).

> and then still gets all his potential end users denied from him [...] How is that behavior on the part of the denier justifiable?

Thanks to this nice formulation above I can finally see the vast ocean of average users eagerly waiting for systemd to be enabled by default (since they are not technical enough to enable it by themselves). They were all hidden by this single guy slowing down the whole process for selfish reasons.

Fedora defers systemd to F15

Posted Sep 15, 2010 22:52 UTC (Wed) by rgmoore (✭ supporter ✭, #75) [Link]

The fact is that people do care about having shiny new things, really quite a lot.

People care different amounts about different shiny new things. They may care deeply about one set, not at all about a second set, and prefer the tried and true for a third set. But different users have different use cases, which means they disagree about which packages desperately need to be updated to include new features and which should be left alone to promote stability. So there may be just 5% of users who want the latest package for Program X and 5% who want the latest package for Program Y, but they're not necessarily the same 5%.

That means that a give user may wind up using a cutting edge distro even if they'd prefer that most of the software in it remain stable. As long as there are a few packages where they want/need the latest features, it may make sense to go with Fedora over CentOS. It just depends on whether it's harder to get the latest version of their program of interest working on CentOS or to deal with the instability of Fedora.

Fedora defers systemd to F15

Posted Sep 21, 2010 10:37 UTC (Tue) by nye (subscriber, #51576) [Link]

>If that were anywhere near true then 95% of users would be using something like CentOS

Or Windows

>The fact is that people do care about having shiny new things, really quite a lot.

I guess that explains why Windows has such a tiny share of the market.

Fedora defers systemd to F15

Posted Sep 27, 2010 15:57 UTC (Mon) by topher (guest, #2223) [Link]

You're right, people love shiny new features.

However, systemd doesn't qualify in any way for that. People don't *see* systemd unless/until it breaks. Hence, they don't care.

The kind of shiny new features people want, and are willing to run Fedora for, are overwhelmingly in the desktop area, or the rare latest-and-greatest server daemon. 99.9% of them couldn't care less about what software is used to handle init.