Calling for a new openSUSE development model

Posted Jun 14, 2012 20:57 UTC (Thu) by drag (guest, #31333)
In reply to: Calling for a new openSUSE development model by hadrons123
Parent article: Calling for a new openSUSE development model

Having deadlines is important if you are planning on doing actual releases because it reduces the number of shenanigans that developers can get away with. If they see the release slipping because of some other developer then they will think 'hey, I can just take the time to pile on some new features or bump up the revision to the latest release from upstream'. This inevitably leads to a sort of mudslide of changes that pile up and you end up with nothing but continuous delays.

Deadlines gives other people justification to reject changes and roll back features that won't make it on time. This has worked out very well for distributions like Fedora which see relatively little delays each time they do a release.

Now deadlines don't make sense if you are doing a rolling release. OpenSuse could do that, but it's going to cause headaches for users unless they stick to a strict policies regarding binary compatibility and such things.

This allows you to continuously feature creep, but if you want to make significant changes (like move from old init to systemctl) then it makes it extremely difficult as regularly scheduled updates used by regular users WILL break systems.

To me the solution isn't messing with deadlines or release styles.. it's something organizational and something that has to do with management style that needs to change. It points to a systemic organizational issue caused by management and thus any fix has to be addressed there.

They have a limited pool of talent and resources. They need to figure out how to either have their organization do less work so they can concentrate on quality control more or figure out how to get people to 'work smarter' and have quality control integrated naturally into the build process.

My personal choice would be to simply latch on to other distributions and 'steal' their work (probably Fedora) so that OpenSuse can concentrate on differentiate features that matter to users and be able to spend more time on quality control.

That way they can improve the status quo without having to spend additional resources.

That is 'work smarter'.

to post comments

Calling for a new openSUSE development model

Posted Jun 14, 2012 22:53 UTC (Thu) by Wol (subscriber, #4433) [Link] (6 responses)

Unless, couldn't you sort-of copy Debian?

Let's say you do six-month releases, then you roll for five like Sid or Cooker, then you do a feature freeze and a mad one-month bug-fest. Then you could "name and shame" :-) if people aren't fixing blocking bugs, and anything that's not a bug fix is simply "come back for the next release".

I know it's not necessarily the best solution, but it's a bit of an obvious one. Says me who runs gentoo :-) but SuSE is my distro of choice when I'm supporting other people.

Cheers,
Wol

Calling for a new openSUSE development model

Posted Jun 15, 2012 0:20 UTC (Fri) by drag (guest, #31333) [Link] (5 responses)

> Unless, couldn't you sort-of copy Debian?

Debian is infamous for poor release management. This is one of the major reasons for the popularity of Ubuntu. They do good quality when it finally releases, but I don't think it's a good model to follow.

The best idea I can think up is to just piggy back on Fedora and concentrate on developing what makes OpenSuse special.

Calling for a new openSUSE development model

Posted Jun 15, 2012 1:46 UTC (Fri) by miguelzinho (guest, #40535) [Link] (4 responses)

Ubuntu does good quality releases and Debian has poor release management?

I'm sorry but my experience is completely the opposite. I have been screwed many times by buggy Ubuntu packages when using a new release early that I've have adopted a Windows user mindset to wait a few weeks to get updated packages after a release, because they will fix a lot of untested and postponed bugs thanks to the release deadline.

OTOH with Debian I do not recall having any serious problems when using an early release at all.

Calling for a new openSUSE development model

Posted Jun 15, 2012 7:26 UTC (Fri) by seyman (subscriber, #1172) [Link] (3 responses)

> Ubuntu does good quality releases and Debian has poor release management?

I believe drag's "They do good quality when it finally releases" comment is aimed at Debian, not Ubuntu.

> OTOH with Debian I do not recall having any serious problems when using an early release at all.

One Debian release (sarge? etch?) was respinned within 24hours of its release. Sarge was delayed time and time again leaving people who needed stability stuck on etch for years.

Calling for a new openSUSE development model

Posted Jun 15, 2012 13:12 UTC (Fri) by drag (guest, #31333) [Link] (1 responses)

> I believe drag's "They do good quality when it finally releases" comment is aimed at Debian, not Ubuntu.

Yes. That is correct.

> One Debian release (sarge? etch?) was respinned within 24hours of its release. Sarge was delayed time and time again leaving people who needed stability stuck on etch for years.

I've used Debian for many years and enjoyed taking advantage of the package management system to do some very crazy stuff. It's nice and flexible. Very useful.

But in terms of release management is that the timing is undependable. It's very difficult to use Debian if you must coordinated with other people in a large infrastructure when the best estimation you can come up with in terms of a release is 'well, maybe next year'.

Calling for a new openSUSE development model

Posted Jun 15, 2012 13:18 UTC (Fri) by Cyberax (✭ supporter ✭, #52523) [Link]

The last several releases were fairly predictable (about 1 release every 2 years). It's the same with RHEL, btw.

Calling for a new openSUSE development model

Posted Jul 4, 2012 21:48 UTC (Wed) by zack (subscriber, #7062) [Link]

> Sarge was delayed time and time again leaving people who needed stability stuck on etch for years.

That was 2005, 7 years ago. I wonder how long a Free Software distro should be held accountable for something like that.

Now look at http://en.wikipedia.org/wiki/Debian#Release_history and take the average of release cycle duration for releases since then. It's 22.6 months, with a very low variance, and it's been like that for 7 years.

That might not match the definition of "time-based release", but it's pretty reliable if you ask me.

Calling for a new openSUSE development model

Posted Jun 14, 2012 23:23 UTC (Thu) by Cyberax (✭ supporter ✭, #52523) [Link] (12 responses)

Do people actually use OpenSuSE like Debian (upgrading only when new release comes)? I've yet to see this.

(Of course, I'd prefer if ALL but one or two desktop distros die out)

Calling for a new openSUSE development model

Posted Jun 15, 2012 10:18 UTC (Fri) by jengelh (subscriber, #33263) [Link] (11 responses)

In the enterprise realm, such compaction has already occurred. It's RH and SUSE that fill in the two spots.

Calling for a new openSUSE development model

Posted Jun 15, 2012 11:52 UTC (Fri) by Cyberax (✭ supporter ✭, #52523) [Link] (10 responses)

Not really. CentOS (well, it's just another name for RHEL, but still) and Debian (also known as Ubuntu) are very much alive and well.

And on Desktop we have Ubuntu, Fedora, SuSE and sometimes RHEL.

Calling for a new openSUSE development model

Posted Jun 17, 2012 0:11 UTC (Sun) by cmccabe (guest, #60281) [Link] (9 responses)

It depends on what you mean by "enterprise."

If you mean web companies and software companies, then yes, there is quite a bit of Ubuntu. Very occasionally you'll see Debian or even something like Gentoo. Non-technology companies tend to stick to Red Hat 5 or 6, at least here in the U.S.

It's easy to forget when you work in software, but software is not a very large part of the overall economy. Sectors like oil, bulk chemicals, healthcare, and so on completely dwarf software and those guys have computers too.

So if you love monocultures, come over and work on enterprise software! You'll also get the fun of dealing with 5-year old system software and its bugs. RHEL 5 is very much alive in the enterprise space.

Calling for a new openSUSE development model

Posted Jun 17, 2012 8:46 UTC (Sun) by k8to (guest, #15413) [Link] (8 responses)

Can confirm.
Over the last 3 years I've seen an increasing enterprise consolidation around Red Hat Enterprise Linux 5. RHEL 6 is now gaining but still a minority. Of course the real majority is CentOS but I consider it a variant of same.

Ubuntu, SuSE, Debian are all far and away second fiddle. SuSE has lost ground, I'd say, while Ubuntu has held steadish, so that they're both now at a very rare-to-hear-about-them point.

Personally I'm happy SuSE is waning because of the crazy shenanigans they've pulled several times now in libc symbol binding games. SuSE is starting to feel like a place where you should only run software compiled by SuSE or yourself. Third party binaries are a minefield.

Calling for a new openSUSE development model

Posted Jun 17, 2012 9:53 UTC (Sun) by jengelh (subscriber, #33263) [Link] (7 responses)

>the crazy shenanigans they've pulled several times now in libc symbol binding games.

Can you actually subtantiate that claim on a /technical/ basis or is this just the typical "I hate X, gonna move to Y"-type rant from a spoiled user?

Calling for a new openSUSE development model

Posted Jun 18, 2012 3:51 UTC (Mon) by k8to (guest, #15413) [Link] (6 responses)

Your level of hostility is off the charts.

Suse 10.1 through 10.4 specifically bound the resolver library tightly by modifying the symbol table with the system allocator to work around firefox bugs. This meant that any executable using an alternate allocator on these versions would end up trying to free memory with the custom allocator which wasn't allocated with it. A crash at this point was about the best outcome you could hope for.

The bug is documented and filed.

There are others I remember less well.

Take your baseless presumptions and go home.

Calling for a new openSUSE development model

Posted Jun 18, 2012 3:53 UTC (Mon) by k8to (guest, #15413) [Link]

Basically this is "you're a liar unless you do all my homework for me." And I consider it trolling.

Calling for a new openSUSE development model

Posted Jun 18, 2012 9:42 UTC (Mon) by jengelh (subscriber, #33263) [Link] (3 responses)

>Suse 10.1 through 10.4

My point is that blaming something today for issues it has had in the past is unjustified. (Just like certain political developments in the real world.)

Calling for a new openSUSE development model

Posted Jun 18, 2012 9:52 UTC (Mon) by k8to (guest, #15413) [Link] (1 responses)

I'm not done debugging the current problem that only shows up on SLES and no other linux distribution that is resulting in memory corruption completely at random.

Calling for a new openSUSE development model

Posted Jun 18, 2012 9:53 UTC (Mon) by k8to (guest, #15413) [Link]

Meanwhile, we do have customers who run those versions, so it's not exactly the past.

Calling for a new openSUSE development model

Posted Jun 18, 2012 15:55 UTC (Mon) by nevyn (guest, #33129) [Link]

> My point is that blaming something today for issues it has had in
> the past is unjustified.

So much sarcasm ... so little time. I shall save my energy and just highlight your words.

Calling for a new openSUSE development model

Posted Jul 8, 2012 1:21 UTC (Sun) by cmccabe (guest, #60281) [Link]

I actually like using SuSE on my personal desktop, so I'm sad to hear about these 10.x problems. Do you have a bug number? My Google skills are weak today, apparently. The number of people using SuSE may be small, but it's not zero (for us), and I hope I don't end up having to debug something like this in the future.