Not logged in
Log in now
Create an account
Subscribe to LWN
Pencil, Pencil, and Pencil
Dividing the Linux desktop
LWN.net Weekly Edition for June 13, 2013
A report from pgCon 2013
Little things that matter in language design
to break useful functionality without even waiting for a major release.
For example, session management was deliberately broken in 2.24. There was no plan to replace the feature. The antique code was simply deemed too ugly to exist.
Fedora happily includes this less-featureful version because, just like KDE 4.0, it's the latest.
Fedora with GNOME is no better
Posted Jan 29, 2009 17:29 UTC (Thu) by jspaleta (subscriber, #50639)
Fedora just happened to get there first...but other distributors are making the same choice to ship the newer gnome-session even with this regression in place. Maybe because other pieces of gnome are starting to make use of features in the new gnome-session design and simply keeping an older gnome-session with have other impacts on the overall integrated desktop.
These comments really speak to the competing expectations projected onto the people working at the distribution level.
Very few distributors promise a feature regression free experience across the distribution release boundary. Why? Because very few of the upstream projects themselves promise a regression free experience across their release boundaries, even minor releases.
But as users that seem to be what we expect from distributors. We try a distribution, we like how the pieces work together and stick with it, and we want it to work that way..forever...regardless of what design decisions are being made in the upstream projects to support backwards compatibility or legacy behaviour. I think its a logic fallacy to expect distributors to do a better job at being upstream developers than the upstream developers themselves.
Fedora certainly does not come anywhere close to even hinting at an implied promise of a feature regression free experience across the Fedora release boundary. Fedora self-describes itself as leading edge, and each release is focused on integrating the current upstream project offerings.
Posted Jan 29, 2009 22:06 UTC (Thu) by omez (guest, #6904)
Posted Jan 29, 2009 22:38 UTC (Thu) by jspaleta (subscriber, #50639)
You are free to call it upstream development models amateur release strategies, but its not going to change how they do releases. A massive injection of additional contributing manpower into upstream projects could have an impact however.
Fedora says the same sort of thing being said in those Ubuntu tickets, except we try to put the stress on contributing manpower to upstream as much as possible. Manpower is limited at every level.
Posted Jan 30, 2009 21:49 UTC (Fri) by dkite (guest, #4577)
Insert upstream projects into that statement.
This highlights a fundamental problem with the desktop.
If you look at server applications, for the most part when they are stable, they cover the use patterns of many, and people are content to stick with what works. New feature requirements are there, but the pace of implementation is slower. The software usage corresponds more to how a desktop application uses a library. Stability stability stability.
The desktop is much different. If I run across some new gizmo, I may be the one informing the developer of the need. And I want it right now. Or I am anxiously awaiting the ability to do something that another platform does with ease. I will use code written yesterday for that reason.
A stable LTS release is useless for the desktop unless there is a very limited usage pattern.
Of course, when I am that close to the edge, things will break.
It isn't a management issue. It is simply that the desktop isn't done, and every improvement is important. Reality forces developers to throw stuff away from time to time to get to the next level of capability.
And the very odd thing is that for a number of years, I've been running KDE from svn. Build every few days. Other than the very early KDE4 time, it works very well.
Posted Jan 30, 2009 22:23 UTC (Fri) by jspaleta (subscriber, #50639)
I know that, you know that... and I know I know what the implications of that are. And I now know that you know as well. Or I think I do.
Anyways.... having the two of us understand what the reality of the desktop development landscape is lets us set our expectations on progress/regressions accordingly. But that doesn't automatically translate into a widespread understanding. I think the expectation mismatch between the user culture and the developer culture is real and if there's someone with some ideas on how to close that expectation gap in the future... I'm all ears.
Posted Jan 31, 2009 0:46 UTC (Sat) by dkite (guest, #4577)
Posted Jan 31, 2009 5:53 UTC (Sat) by malor (subscriber, #2973)
It's really that simple. Call a product what it actually is, and it's amazing how much better consumers will like it.
Posted Jan 31, 2009 14:08 UTC (Sat) by kragil (guest, #34373)
4.0 = 4.0 alpha
4.1 = 4.0 beta
4.2 = 4.0 RC
4.3 = 4.0
I will betcha that if they went down that road the 4.0 we would get in 6 months would have more unfinished blue sky features and frameworks and would have been less usable than the real 4.1. (again look at E17)
And I won't even talk about the number of ported apps.
Net result: 1 year+ wasted.
Release EARLY release OFTEN.
4.0 did work and was OK to release .. the "mostly erects the pillars of KDE for developers and upcoming releases" part is the only thing that should have been in big letters in every announcement and blog post and Fedora should have been more conservative.
Posted Jan 31, 2009 21:08 UTC (Sat) by jospoortvliet (subscriber, #33164)
For example. NVidia would've continued to ignore our complaints about the
lack of proper XRENDER acceleration. The "better 4.0 release" would
therefor still have had horrible performance and corrupted systemtray
icons. Even 4.2 suffers, on some installs, from those issues - X.org and
the driver developers haven't been able to fix that in the last year. The
same goes for more pieces of technology, major functionality in Qt being
one of them.
Another reason to release was to get more developers. Because as you might
know, a FOSS project depends on developers, and doesn't really care about
users. Users only eat time away from the actually useful stuff:
development. The 4.0 release succeeded in this regard: over 300 new
developers in the last 12 months.
There are more reasons for releasing 4.0 when we did, but I hope I managed
to make clear that there was more than "the end users will love this". End
users were simply not the target for this release, we did it for other,
equally (if not more) important reasons.
The only thing we imho could've done better is the release announcement
itself (compare it with the 4.1 announcement and see that we DO learn).
Pretty much every other piece of communication was handled well.
Posted Feb 1, 2009 0:30 UTC (Sun) by malor (subscriber, #2973)
You have explicitly said you called it 4.0 to get more testing; I'd say that users were exactly the target for the release. You wanted them to start yelling at other teams to fix the bugs you wanted fixed. You deliberately inflicted pain on your userbase so that they would browbeat other teams into doing things the way you wanted.
But, guess what? You're writing a desktop. If you're not completely focused on users, you're doing it wrong. You deliberately inflicted pain on people to make your own lives easier, and you did it through deception. And you seem to believe that it was even worth it, while GNOME keeps eating you. You're losing, and you are losing because the people who vote, the users, aren't the people you're really trying to help.
Remember: Linux users are generally developers of other projects. Their time is worth at least as much as yours.
we did it for other,equally (if not more) important reasons.
You're focused on your own team's needs intead of the needs of the broader public. This is damaging you, badly. If you don't want to focus on users first, then pick a different type of project.
Posted Feb 1, 2009 11:47 UTC (Sun) by jospoortvliet (subscriber, #33164)
You can look at it either way - the short term annoyances were worth the trouble. I honestly believe time will prove us right.
Thanks for KDE 3.5.10!
Posted Feb 1, 2009 13:44 UTC (Sun) by mgb (guest, #3226)
Fortunately there is a good, stable KDE in 2009. It's 3.5.10.
Many thanks to the subset of KDE developers who are still supporting their stable 3.x series, even while experimenting with new ideas (some good some bad) in the 4.x development series.
Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds