Welcome to the Arch Linux Newsletter. This document attempts to give you an
"at a glance" look at the world of Arch Linux.
This week we have had a minor rejig of the contents, placing all of the stats at the bottom of the letter and all the exciting stuff at the top. We also have our first interview, so many thanks to Jens for setting that up.
This time again, we have no front page news. It seems the newsletter contributors are catching up with the developers.
foldingathome pkg to [community].
pacman -S foldingathomepacman -S kernel26archck. kernel26cko will not be updated.This is an interview with Ions founder Tuomo Valkonen. You can find more info about
Ion in this wikipage. We are very grateful
that Tuomo took the time to answer our questions with such long
and great answers. :) I'm swedish, Tuomo is finnish and the language is english
so please have some mercy on us.
Can you give us a brief description of what Ion is all about?
Ion is one leg of an odyssey in search of better graphical user
interface paradigms.
What's new in Ion3 compared to the previous version?
Ion3 is not yet finished, and there may still be some significant
improvements to be written, but at the moment the most notable improvements
possibly are simplified configuration routines that external programs should
be able to understand too, a nice statusbar that adapts to workspace layout
instead of wasting a whole border of the screen, floating splits and the
scratchpad. The scratchpad
is a frame toggleable visible with Mod1+space by
default, which is modelled after the console of many FPS games. Floating
splits allow frames to partially overlap in a controlled and thus usable
manner. This feature should improve Ion's usability with side-by-side
applications that benefit from slightly bigger windows than a strict split
would allow for.
Other new features and improvements include the "pane workspace" that is
an attempt to make window management even more automatic. It, however,
still needs more work, especially with heuristics and other ways
to classify windows that are needed due to non-existent application
support for indicating the type of a window. Obviously there are
numerious minor improvements and fixes as well, and the scripts
repository
is growing.
A few people have asked me why Ion hasn't got a special
configuration file, possibly XML based, so I forward the question over to
you. Will there be a configuration file or will it stay with Lua only?
There is a configuration file. Many infact, they just happen to be Lua
scripts. And no, there are no plans to support configuration files in a
non-scripting language format. Even if there were, the format would never
ever be XML. It's not meant for human consumption, and configuration files
should be human-readable. Something like .INI is a rather robust and easily
readable format, and therefore much better for configuration files. For
documents a TeX-like syntax is quite ideal, although one would hope for a
non-Turing complete lookalike with better separation of structure and
layout.
Back in the topic of configuration files, robustness is indeed the
biggest problem of using a scripting language for configuration. Make one
syntax error and the whole file is unusable, and this can be a big problem
with newbies. The reason for not supporting a more robust configuration file
format is that supporting both a scripting language -- and people do want
scripting -- and configuration files in a different format makes things more
complicated than necessary and brings extra work maintaining two redundant
ways of doing the same thing. With the changes in Ion3 it should however be
possible to modify the configuration files with a helpful tool instead of
only directly editing them, and there has been little work towards such a
tool, although not much recently I'm afraid.
Are floating windows all bad?
If you include "floating splits" as introduced in Ion3 as floating windows,
then the answer is clearly negative. If you restrict the definition to
freely floating windows of the conventional WIMP desktop style, then the
answer is pretty much positive. The biggest problem is that such a window
management model doesn't scale at all. It can be used to a degree with maybe
a handful of windows when cycling through all of them isn't a big task.
Something like Exposé may increase this number a little, but even it should
run into problems once the windows are too many to make to draw a good
scaled-down picture of, or when they are almost identical. Locating a needle
in a haystack is slow, and this is one of the major problems of the whole WIMP
paradigm. When you have running a few dozen xterms like I do, you can't
cycle through all of them effectively, and you can't quickly or at all pick
them from a list of images nor titles that might be nearly identical. Until
computers can read our thoughts, what you need is a way to keep the windows
organised in such a manner that it helps remembering where the window was.
PWM was a first step in providing such a way to move between windows
attached to the same frame. Ion provides ways to find the right frame.
Ion does, however, still have a scalability problem. It doesn't easily scale
to dozens of small windows that should be used almost constantly. Mostly
this means the toolboxes of GIMP and other such programs. But neither does
the intended environment scale to this; one needs to keep constantly
arranging the windows. I think this kind of UI design is inherently bad, and
everything needed to edit a document should be found in the document's
window. (Compare Sodipodi to Inkscape.) Yet, the "pane workspaces" try to
address this problem by automatically putting such toolboxes in a particular
pane, document windows in another and so on. As already mentioned, there's,
however, much work still to be done with it to make it really work.
What's good and what's bad in todays applications GUI designs?
I can't think of much good to say about them, and many things in the FOSS
world are getting worse infact. More like windows. What I think of the Gnome
file chooser isn't fit for printing. To list some general annoyances of GUI
software of today: zillions of dialogs and toolboxes popping up everywhere,
applications insisting on their own weird skinning instead of the one
provided by the toolkit when using one, applications trying to be their own
window manager, widgets laid out in a complex fashion such that keyboard
movement between them is totally unpredictable, practically unconfigurable
bindings with the defaults being very windowzy and using modifiers in an
inconsistent manner, bad keyboard support in general, and so on.
Good keyboard support does not mean just shortcuts to everything, it means
designing the application in such a manner that one can effectively use it
with just a few shortcuts, instead of by memorising every single escape meta
alt control shift combination imaginable. I'd dare to claim that commands
are infact much easier to memorise, and not all that much slower to type.
Infact, there's one GUI app that has got something right. It's TeXmacs. It's
graphical, and it can interpret LaTeX-like commands in the input stream.
Yay! I'd prefer a more "wysiwym" instead of "wysiwyg" approach, though, but
that's almost as good as wysiwyg can ever get.
Have you been following the development of other DEs/WMs? If so,
what's hot and what's not? Have you tried WMII?
I don't much follow what's going on in the DE world. Whatever happens, I'm
sure it sucks from the point of view of trying to write an alternative
window manager and promote the concept of subjective usability, the Official
Truth saying that usability is absolute and means "noobs aren't scared of it
according to how we think noobs think of things". I think it is wise to be
scared whenever there's news from the Gnome camp.
I've tried WMI(I) a couple of times, and it does have some good points, but
unfortunately at least those snapshots I've tried have been rather flaky.
What do you think about all the new X11 stuff?
The only thing that made upgrading to Xorg worth it is working Xrandr
rotation. The XComposite extension could be useful for something, although
its primary intended uses are mostly worthless gimmicky. As for most of the
other new stuff, I don't like needles being sticked into my eyes after every
major upgrade or new install by apps using ugly antialiased fonts and
refusing the use good old X fonts at all. (I had to spend an hour or so a
few weeks ago to find out how to get the beautiful old X helvetica in
firefox -- unblurred of course. It wasn't a small task.)
Are there any plans made for a Cairo based Ion version?
I did think about writing a Cairo drawing engine module once, but any plans
have been abandoned. I don't like having to do extra work to get unblurred
fonts.
Are you involved in any other projects?
Well, I'm not at the moment actively working on any other free software
project. I'm trying to concentrate on finishing Ion3. One day "when it is
ready", after which it is time for something else for a change. I did write
Riot (Riot is an Information Organisation Tool)
last autumn, and have some plans on further work on it. A bit related
project to which I've contributed some code to and may do some further work
on is Yi. Then there's of course
Vis, but it is still at the vapourware stage, and likely to remain there for
quite some time being a huge undertaking that I'm not ready to take at this
point, trying to finish a particular project that grew bigger than I ever
expected :), and having started working on another big project, namely PhD
studies. Vis indeed stands for Vapourware Interface Synthesiser, and it's
another leg of the odyssey I mentioned. The write-up available from the
Vis home page tries to explain how it would be
possible to write programs in a very interface-agnostic way, so that the
user could ultimately control the kind of interface programs had while also
relieving the programmer of the burden that modern GUI toolkits are to use.
Even if I don't, I do hope someone else gets around to write something like Vis
and it becoming a standard way of constructing user interfaces.
Ok, last question. What's your favourite distro? Have you tried Arch?
I've used Debian since about 1998 and have had no reason to try anything
else. If it works, don't try to fix it.
Editors note: You may publish this interview separately (screenshot also) free of charge if you
add a note that it was originally published in the Archlinux Newsletter.
Q: What is the difference between AUR, COMMUNITY, and UNSUPPORTED?
A: AUR(Arch User Repository) is the name of the whole new system which is maintained by the TUs(Trusted Users).
It lets anybody that signs up to upload PKGBUILDs to UNSUPPORTED but only TUs have access
to COMMUNITY which is a binary repo that you can use with Pacman.
To Participate, visit:
http://www.archlinux.org/~simo/archstats
Number of registered systems: 1266
Longest recorded uptime: 496 days, 9 hours, 22 minutes, 38 seconds.
Average uptime: 7 days, 3 hours, 51 minutes, 24 seconds.
Least packages installed on a system: 44
Average installed packages: 322
Most packages installed on a system: 1624
Lowest Bogomips: 388.09
Average Bogomips: 3279.59
Highest Bogomips: 7487.48
To Participate, visit: http://setiathome.berkeley.edu/team_display.php?teamid=116975
Members: 7
Total credit: 5,532.84
Arch Linux Team Page
Team Number: 45032
Members: 3
Score: 8825
Ranking: 5855 of 39519
That's it for this time folks. If you have any opinions on the newsletter or have some
things you wanna add, just send us a mail and we'll look into it.
Very best regards / Team Arch