Proposal: include gnome in xubuntu

Jani Monoses jani at ubuntu.com
Thu Feb 14 21:29:39 GMT 2008


> After some more reflection I think I see why I'm not happy with the changes.
> In fact it's not really a matter of gnome or not gnome (not entirely).

Great!

> Having the best softwares in a distribution is a good thing, no doubt about
> that. But I feel that the way it's been done since Gutsy (read, add
> applications, no matter where they come from) is probably not what I expect
> from a distribution built by a community. The goal seems now to provide a
> fully featured distribution, in the easiest and fastest way possible for
> devs. This is what Ubuntu wants to provide, and what users expect Ubuntu to
> be, this is the way a distro which aims to be the leader has to proceed (and
> thus use a fully featured DE as base). Is that what we want from Xubuntu as
> developers (and as users too)? I don't want of this, and I find way more

Actually this is what I wanted from Xubuntu from start. A distro as easy 
and as complete as possible for lower powered hardware. The emphasis, 
being part of the Ubuntu family was not on the origin of the software 
but on how useful it is to users, especially new ones. Xfce was the 
ideal choice for a lighter desktop.

Initially I thought that using only Xfce and non-GNOME apps would do. 
Two years ago Xfce was about to release the 4.4 release, vastly superior 
  to the previous 4.2. They had momentum and there was a lot of 
enthusiasm around the project. Two years later there have been no 
serious improvements except bugfixes (GNOME had 4 stable releases in the 
meantime, each better than the previous, with small but steady and clear 
improvements) The planned features that would have matched GNOME and 
that were talked about then (smb support, better desktop icon handling, 
XDG compliant menu, file search, etc) and new GTK only apps that would 
be lighter but just as good as the GNOME ones never materialized. That 
is not a surprise: it takes a lot of effort and more than 5-6 developers 
working in their spare time.

I contributed as my time and skills allowed back then by porting several 
  panel applets to the 4.4 panel API, fixing bugs in xfce-session and 
xfdesktop and packaging their svn trunk regularly for dapper.

But for post dapper releases I realized that since we cannot expect from 
  only a handful of devs and us to supply all the nice features that 
would make a user comfortable and keep using Xubuntu we had to look 
elsewhere.
Whoever thinks now that we could start projects to write such apps from 
zero is well meaning but totally unrealistic. It does not work this way,
there is a reason why GNOME and KDE are where they are after 10 years 
and hundreds of contributors.

As long as you use the same tools (C/C++/Gtk) there is simply no way to 
keep up with that amount of features even if you had only coders as 
talented as Benedikt (Thunar author)

> interesting to spend some time on improving the distribution by other means
> than just using what's been done by the others (yes, that's a dev POV, but
> we are on a devel list, right?).

We are devels but I at least am a developer because I want to make 
something for users. I am not needy enough to scratch my own itches :)

And I think we could look at a distro as something we 'distribute' so 
the emphasis is on the final product not on our feelings while making 
it. It should be done with care and enthusiasm but without user 
satisfaction those are just selfish and nothing more.

And I think that it is a good idea actually to use what others have 
done, but also do new and creative things. But IMHO that does not mean 
throwing out  what some people have done and use what another set have.
Because frankly introducing xscreensaver, squeeze, etc is not actually 
rocket science or fun: you still take what others have done but you can 
say that you are a bit original.

I think the whole 'do not copy Ubuntu' idea can backfire. In search of 
identity and originality the users are loosing. They do not know - the 
newbies I mean, the advanced users are not very affected by our default 
package selection - where software comes from upstream, they only see if 
it works or not.


> I do care about users, even though I've been blamed several times already
> for not taking real care. That's a blame I have to accept I guess... :)
> If building an Open Source distribution is just a matter of collecting
> pieces here and there without doing anything else, well I'll have to find
> other things to do in my spare time, this is not interesting.

That unfortunately is what making a distro is most of the time. It seems 
quite boring to me too :) Anyway this is not the case we are arguing 
about, it is not like I said let's reuse gnome apps while Jerome and 
Lionel had been hacking on alternative gtk only apps and having fun. 
Their actions too were picking existing apps. Different upstreams but 
similar processes. Actually more effort because they are not shared with 
Ubuntu.

There are actually fun things to do, look at Ubnutu devs doing extra 
apps besides what upstream does. That would be great in Xubuntu too.
But adding features and things that matter to users, not just working 
for fun without a clear goal.

> 
> This might sound silly, but why not take more time between the releases? The

It sounds very unrealistic to me. Since we share a lot with the rest of 
the Ubuntus we could not have a different schedule without bothering the 
other Ubuntu devs (freezes, base system, kernels, ISO images) or having 
enough manpower to do it ourselves.

A different release schedule is overkill.
The only reason we could need a different schedule that I see is if Xfce 
had a schedule. Then we could say we make 9 months or whatever schedule 
to fit theirs. But even that is unrealistic. Xfce is orders of magnitude 
smaller and more flexible than Ubuntu, if anyone, they should adapt to 
this schedule.

> 6 months based schedule is a good point for Ubuntu, is it really useful for
> Xubuntu? Jérôme has done a very good work on the bug triaging, and managed
> to build the best relation I've seen since the birth of Xubuntu with the

that is great, congrats Jerome!

> Xfce developers. How could Xubuntu have a good relationship with upstream
> when the only feedback is "This xfce application is not as good as the gnome
> equivalent, we won't use it until it's better" ? Taking the time to test the

I am not sure they care. I still see a clear distinction between Xfce 
core (stable, fast, quality code) and xfce extras and other apps. They 
have no GNOME-like scrutinizing process by which to accept a project 
only when it is mature enough.

So whenever we talk about upstream we talk about two sets of apps and 
most of the time to sets of developers with whom we should have 
different relations and not treat them the same.

Seeing enough of GNOME and Xfce in the past two years I am convinced 
many of the Xfce extras and apps are either made only for fun and 
learning or for personal use. If done right these usually mean user 
acceptance and success too but it is far from being a rule.

> applications, make them crash, report and fix bugs is certainly a better
> way, isn't it? But yes, it takes time. Does Xubuntu have deadlines? Does it
> have something to demonstrate to some customers or investors?

It has something to demonstrate to users. You don't mean that if not a 
commercial project we should drop deadlines? Clear schedules is one of 
the best things a project (or any activity for that matter) could have.

'Ready when it's ready' is annoying unless the outcome is sure to be 
awesome and make up for all the wait.

And yes, as you say it takes time that could be better spent on 
improving what we have. Starting from zero with no clear benefits (and 
this is the main issue) is less optimal.

> 
> Maybe I didn't realize what really is Xubuntu after all. Please tell me if
> this is the case.

True, I think I have never thought clearly about it or formulated and 
noone has. To me it is Ubuntu (users, users, users) for lighter 
desktops. Only Xfce qualifies as the other light WMs are not firendly 
enough. Of course to others Xubuntu may mean something else.

And the fact that it significantly overlaps with Ubuntu still allows it 
to work on lower hardware. I see a certain beauty in doing less work 
with more benefit, that doing a lot of work for work's sake and fun but 
with no visible gains.

I think that having Xubuntu and Ubuntu share apps allows us to spend 
time in improving both at the same time. I care about free software 
primarily not about Xfce or GNOME so for me there are no issues that 
make me put such a decision before the user PoV.

For distributions that are meant to promote Xfce or certain software 
there are other fine alternatives, which have a much better chance of 
being lightweight.

> 
> Quick note on the technical side, I'm working on a Xubuntu derivative for a
> very low end machine (256 Mo RAM, 700Mhz equivalent processor) since one
> year, adding gnome libs has a real impact on the performances (but no, I
> don't have numbers to show, so it probably means nothing...).

You do not have to have numbers, it is an uncontested fact that linking 
to gnome libs makes apps use more RAM and start slower. This is what I 
have been arguing about with some GNOME developers for over two years :) 
and this is why I have sent patches to remove those to at least 10 apps 
already.

We all feel that after 6.06 Xubuntu got slower. *But this is not due to 
libgnome linking apps* This is the main fallacy we seem to get stuck on 
and based on which decisions are made.

The apps that link to extra libraries be it libgnome or other are 
bloating if they are run all the time: panel applets or desktop daemons 
are good examples.

So while for evince it helps to have it gtk only it only affects 
performance while launching (say 200 ms extra) and uses up 1-2 Mb of RAM 
more.

But only when these add up in long running apps (update manager, network 
manager) does it affect resource usage permanently.

And even so, we now have things in the base system that are slow, and 
GTK got slower, font rendering got slower and there may be other factors 
as well.

So throwing out an app that stays put in the menu and does not cause 
bloat with another will only give us the illusion of lightness. You may 
feel light at heart knowing there's no gnome app there but you'd be 
wrong about bloat :)

Back in 6.06 the drastic no libgnome policy was helping with CD size 
too. But once we did not want to be absolutely spartan - no update 
manager, no cd burn app, no stable archive manager - and we let at least 
one such app in the cost for the rest was 0.

And no, such apps were not introduced as trojan horses in order to 
slowly sneak in the whole of gnome but
1) we needed at least a few of those apps and
2) once we had them there is no loss by using the others.

Had anybody patched out libgnome from those essential ones or had gtk 
only alternatives we'd have chosen those. Fun projects for whoever wants 
to take them, but only shows that things are not as easy to make.


thanks for reading so far, I hope this clarifies my view on Xubuntu and 
explains some of my actions. I could have explained them earlier if asked :)

I'd love to hear other's views, if possible with similar amounts of 
explanations :)

Jani




More information about the xubuntu-devel mailing list