Hello!
I promised on IRC to write a bit about my experiences with installing
Foresight, but it apparently took me a while to finally get around to
it. Sorry for that.
To make it clear where I'm coming from, I got interested in Foresight
when I read the interview with Ken on pro-linux.de, mainly because of
his views on Usability. I have been spoiled by Fedora and Ubuntu
lately, so I'm very happy that Foresight shapes up to become another
user friendly and polished GNOME desktop. Right now I am using Ubuntu,
but Foresight is happily sitting on another partition. I am not sure
yet if I'll totally make the switch, this depends to the biggest part
on whether I agree with the goals of Foresight. More about this later.
Alright, so here are my installation experiences in short-form:
- The installer theme seems to be broken, I assume it's using the Clearlooks gtkrc but not the actual engine.
- Overall the install went very smooth. I like Anaconda and it brought
back memories of installing RH/Fedora. ;) The Ubuntu installer is a
little simpler at places (especially network config) and asks less
questions.
- It asked me for a root password, although Foresight seems to use sudo
for almost everything. This is probably arguable, but IMO one of the
main user-visible advantage of using sudo is, that the user doesn't
have to remember another password (or learn what a "root" is). Ubuntu
leaves out this step but immediately sets up an administrative user. I
haven't used my root password once yet, despite adding a user at first
because the first boot wizard was broken.
- The first boot wizard was broken. :P It didn't appear at first, but
later it suddenly popped up after logging out of GNOME, so maybe the
issue was fixed meanwhile.
- While installing the NVidia driver was painless, I'm always a bit sad
when I have to do this manually, because that's pretty much the biggest
issues which makes me think twice before recommending a Linux
distribution to other experienced PC users, like gamers. Ubuntu makes
an exception for non-free drivers (but they don't support them), and I
think that's a reasonable thing to do at the current situation. So I'm
wondering, if there are any plans to package those drivers and make it
simple to enable them? There is nothing about the license which would
make it impossible or legally questionable to include them, they are
just not free software.
Desktop Impressions:
- There is no fancy X cursor. :) There needs to be a fancy X cursor!
Ubuntu uses Jimmac's Industrial cursors and I think they are a great
choice: http://gnome-look.org/content/show.php?content=6550
- Now that the Home and Computer icons can be reached from "Places" and
dragged to the desktop, I really don't think there should be any
non-removable icons on the desktop. This is probably an upstream issue,
but a work-around would be to disable the appropriate gconf keys and
instead put regular links to Home and Computer on the desktop, which
the user can remove (just like those you get after dragging the items
from the Places menu). The trash is IMO much more usable on the panel
anyway, so I would suggest to disable this icon and add the trash
applet to the lower right panel corner by default. Then the user would
entirely own his own desktop and could remove all default icons without
irreversibly losing any functionality.
- The panel had some issues, namely the launchers didn't prelight and
after accidentally clicking on Delete this Panel, it was deleted
without any confirmation. That's totally not correct.
- Sound was totally muted (both master and PCM), so at first I thought
it wouldn't be working. After changing it, it was correctly saved, but
the defaults should be sane.
Applications:
Well, this is what bothers me the most at the moment. My opinion is,
that a great free software desktop should appear as one integrated and
polished piece, not as a random collection of differently branded third
party applications. This is actually what I liked most about Fedora and
most of all Ubuntu. What this means is: Focusing included packages on
tasks, so there is exactly one tool for any job and only including
tools which are totally integrated into the desktop and are on an equal
level of quality (if possible). It also means de-branding applications
and using icons and names which accurately describe the task. Some of
the worst offenders in this regard are Ximian/Novell applications,
which are otherwise excellent of course. For example Beagle right now
is not integrated into the desktop at all. It is certainly fine for
testing, but otherwise it's pretty poor to have a menu item called
"Best", with the icon of a dog. ;) It also needs to be figured out how
this will work together with the current "Search for Files" interface.
A good example for an application which IMO should not be installed for
a typical desktop installation would be GnomeBaker. Mainly because the
actual task of burning CDs is already covered by Nautilus, in an
infinitely more integrated way. We can already burn data CDs by
dragging files in a window and selecting Write to CD, we can burn
entire ISOs with a simple context menu entry, soon we can copy CDs just
as easily, Rhythmbox will allow us to create audio CDs, and so on.
GnomeBaker and other third party tools also apparently use a different
method to detect CD writers, Nautilus is the only burner which happens
to automatically detect my device, which offers a tremendously better
user experience. By mixing and matching applications of such different
quality, I feel that the impression of the entire operating system is
dragged down. Let's look at this from the perspective of a first time
user, every complicated or low-quality tool that is part of the base
install will affect his or her opinion about the desktop, every
inconsistency makes it harder for them to feel at home and confident
about their work. This doesn't mean that apps like GnomeBaker are bad,
they are very cool and useful tools, I just believe that they don't
belong into the desktop, but should be available as optional add-ons
(either from the installation CD or the repositories).
I'm sorry for this long rant about this apparently simple issue, I
don't want to convince you of anything. It's just important for me and
I'm very curious what your opinion about this is.
Phew... ;) Finally, I was very positively surprised by Conary. I have
yet to really understand it, but it's already obvious to me that this
is not just another packaging tool. I had some troubles while upgrading
stuff, I might write more about this later. I was extremely amazed how
easy it was to write my own recipe, especially as I had exactly zero
experience with software packaging. However, I think that from a user's
perspective, it's somewhat awkward right now. The GUI is nice and
simple (apart the rough edges), but looks like it's more appropriate
for system administrators than mere users (but the same is true for
Synaptic and co). I think to make the goodies of package management
accessible to everyone, it should work something like this:
1.) Optional automatic scan for updates, including a nice and simple
GUI to list available updates and initiate the update process,
including nice and friendly feedback.
2.) A tool like gnome-app-install or the Red Hat software install tool,
which allows to easily install or remove groups or single packages of
the distribution. This could for example by used to easily install a
build environment, without installing it by default.
3.) A searchable repository which lists only actually interesting
desktop applications with friendly descriptions, maybe screenshots...
This would probably be similar to "Click and Run" from Linspire, just
without the additional cost. Maybe a simple way to do this would be via
webbrowser integration. Not THAT simple of course.
I understand that this would take some time, but it would be nice to hear your thoughts and plans about this.
Oh well, I probably should split up this message, but now i want to get
it done (meanwhile I got a little tired, so my rambling might become
harder to understand)... The questions I have:
- I couldn't fail to notice some similarities between Conary and
Portage. There would be the use of Python, the syntax of the recipes,
and there even is an "emerge" command and Use flags. I assume that's
not all by accident. :) So, was Conary inspired by Portage? How would
you describe the similarities and differences, other than the obvious
one that Conary is built around binary packages? Are there any plans
for conary emerge to support dependency resolution, so it would become
a real alternative to Portage?
- One of the main advantages of Conary seems to be the distributed
development, but the details are not totally clear to me yet. It's
definitely impressive how Foresight is built upon Specifix, but if I'd
want to factually describe anyone why this wouldn't work as well with
any other packaging format in a way that can actually be understood by
dummies (like me), what would I say? What I (hopefully) understand so
far is, that Conary will update packages only from the branch they were
installed from, is that the main difference or is there more to it?
- What is your opinion about autopackage(.org)? Another packaging
format once again reminds of the question how ISVs are supposed to
distribute their software. Especially small development teams often
have no easy way to build binary packages for countless of
distributions. Autopackage tries to solve this and they are using a
great job, loved by most users and developers for it's usability and
possibilities, but mostly criticized (or met with silence) by the large
distributors. It is strictly designed to complement distribution
package management instead of replacing it. Would you consider
supporting it in a basic way, like pre-installing the support code (so
it does not have to be installed from the internet on first
installation of an autopackage) and making sure that autopackages
integrate fine? They also plan to integrate with native package
managers, for example to install dependencies and to register the
package with the package database. Do you think this would be feasable,
considering the architecture of Conary?
This should do for now. :)
I noticed that the community is still extremely small (quite the
culture shock coming from Ubuntu ;)) and I couldn't find much in terms
of press coverage. The first few times I heard of Foresight, it didn't
seem worth a closer look to me, because it was marketed "just" as an
experimental distribution with up-to-date packages, not as a serious
project. I found the interview with pro-linux.de very interesting and
it totally changed my mind about it. There doesn't seem to be anything
like this interview on an international news site, so I have been
playing with the idea of wrapping up my experiences and eventually some
of your answers :) to a little introductionary article about the
distribution and then submit it to OSNews. Maybe they would publish it,
because they regularly publish articles like this. Maybe that would
interest more people into the project, because I do think that it has
lots of potential. Of course knowing my own lazyness, I can't promise
anything. :)
Bleh... It's frustrating how some short thoughts translate into such a bulk of text. Maybe I just need some sleep. :P
Daniel
