Some reflections on the future of partman [was: Partman]

[Anton Zinoviev <anton@lml.bas.bg>]

Hi!

In general Partman suffers from two things.

First - it is very slow.  I'd say it was designed to be slow because
when I started thinking about it I didn't expect that it will become as
slow as it is now.  Moreover adding additional components to it make it
even slower...

Second - its interface is based entirely on debconf.  This means that
its interface is linear, no button bars, no mouse, no keyboard
shortcuts, no menu bars, etc.

In order to solve the first problem we have two choices:

1. Wait until the computers became faster.  At 2009 when sarge+1 will be
released [;-)] they will be about 5 times faster than now.

2. Rewrite it in some other language.  If we take this approach I would
suggest to package some scripting language (perl, python, guile) as udeb
as making big changes in C programs is a pain.  If I wrote partman in C
there would be no big difference between its first version and its
current version.

In order to solve the second problem we have these choices:

1. Add to cdebconf additional types of questions.  This would make the
interface considerably better, but not as nice as the interface of the
other installers.

2. Do not use debconf, but write directly to the screen.  I'd not love
this approach.

3. See how this problem is solved by SuSE - YaST supports both text and
graphical installs, both have very nice look and as far as I know there
is almost no duplicate code (due to text+graphics).  Notice that
recently (and finally) SuSE released YaST as free software (GPL).

4. The fourth approach is described in the second half of the following,
but now it seams to me fantastic:

http://lists.debian.org/debian-boot/2003/08/msg00015.html

I hope it would be possible to adapt partman to any of these approaches
provided it were faster.

Anton Zinoviev