[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

My doubts about Debian choice





I am still waiting for an official announcement about basing on
Debian.  Omega promised to justify it and he still has not done it.
In fact the announcements I have seen in Linux-Announces or in the
Linux Gazette still speak of a Redhat base.  In my country a law
cannot be enforced until it is published, so due than the Debian
choice has not been published it cannot be considered definitive.  In
addition it was never debated.


First: I am not a Debian user and this time I regret it, because I
would be saying the same but that would not seem advocacy from a
Redhat user.  IMHO when discussing SEUL matters it is _important_ we
put aside why we like in the distribution we use and concentrate _only_
in what pa, ma or little sister would like, on what is good for this
project and still better in what is good for Linux.

Second: Bruce Perens has mailed me the following and authorized me to
quote him despite my warnings than it could used against Debian.  That
gives credit to his honesty.

> One of the reasons we have 6 web servers and no andrew or thot is that we do
> not tell the package maintainers what packages to work on. They package what
> they want. Thus, if nobody is interested enough in andrew or thot, they will
> not get packaged. I would hope that some of the people who are interested in
> ease-of-use would volunteer to package these programs.
> 

In a free project nobody gives orders but it seems in Debian there are
not even suggestions or something like periodic announcements about
"Useful software we would like in Debian".

What I infer of Bruce's post is than the Debian project is so loosely
coordinated than there is a definite possibility than we get only
minimal help if any even if Bruce tells Debian developpers to help
SEUL.  I also doubt of the value for our project of people who have
been developping a distrib since 93 or 94 and despite this in four
years nobody between these two hundred people has shown the slightest
interest in packaging a WYSYWYG Word Processor.


So I will expose here what I feel are the comparative advantages and
drawbacks of Debian against basing on RPMized distributions.  Notice
than there is a part of subjectivity.


Debian's advantages:

-DPKG:  The best package software in Linux world.

-Integration.  A very good point in Debian and I certainly want SEUL
imitate it on this point.

-General quality of the packaging one of the reasons being the
existence of a document defining how to do it right.

-Help:  A good point.  If we get it.


Debian's drawbacks:

-Dselect: Completely brain damaged from an ergonomic view point.  We
simply cannot use it.  Deity is still not released, until then it will
be difficult to build the install part.

-DPKG is not the standard.  By choosing DPKG we cut ourselves and our
users from all the remaining distributions.  That will make difficult
transitions from/to SEUL.  Notice than commercial software uses either
TGZ or RPM format and never AFAIK DPKG.  And "alien" is only good when
you want to try a software not for using it seriously because you lose
too many important features like install scripts.

-The majority of users and specially beginners are using RPM based
distribs.  We will be more useful to them if we exert a positive
influence on the distributions they are using and that will be easier
if we base on an RPM-based distrib.

-LACK OF SOFTWARE.  Yes I know about Debian's volume but despite this
it is poor in END USER software.  Everybody in Debian is very excited
in DPKGing esoteric programming languages or web servers and nobody
has given half a thought in Wysywyg word processors or simple
Databases (NOT Postgres).

-Many packages are not in the shape we want.  In too many cases the
Debian packagers have left to the user to customize them to something
acceptable.  Examples of this are the Debian kernel and many X apps
who come without resources.  Our users cannot customize.



-Advantages of RPM world:

-It is the standard.  That will enable us to use packages from Redhat,
Caldera, Suse and about every other distribution except Unifix,
Slackware and Debian.  In addition there is the redhat-contrib part.

-The distributions using it are targetting corporate users where time
is money.  They are designed to run out of the box.  An example of
this is than RH, Caldera or Suse can be used in any networking role
out of the box while the Debian user gets only basic networking until
he recompiles the kernel.  The "run out of the box" spirit leads to
distributions where the user gets acceptable defaults without doing
anything special.  And that is precisely what we we are aiming at.

-More end user software specially if we also consider what is
available in Suse, Caldera and redhat-contrib.  Andrew _IS_ shipped
with Suse, Caldera _HAS_ a simple to use database and all this can be
downloaded or (in the case of Caldera Lite) got in ultra-cheap CDs.

-The trend in the RPMized distributions is to include features like
hardware auto-detection, modularized sound who does not need be to
recompiled or automatic mounting of floppies and CDs.  All these
features are exactly what is needed in SEUL.  By basing on an RPMized
distribution we will make easier incorporate all the future features
of distributions using RPM and these distributions tend to care more
in user-friendliness than Debian.


Drawbacks:

-The quality of the packaging in redhat-contrib tends to be poor.  I
have renounced to use RPMs from it and I always download SRPMS and
build the RPMS myself.

-The fact than we will be getting software from many sources and in
addition the fact than the RPM world is presently split between glibc
and libc5 will lead us to rebuild the RPMS ourselves.  It will be
necessary to define the development environment in great detail: you
don't get the same GIMP if you have or lack the mpeg libraries.

-RPM is not as good as DPKG.  Its front end GLINT is nice but
underpowered and dangerous.  Fortunately the dangerousness can be
fixed in a few minutes work (I did it) and there are some easy fixes
to make GLINT better.

-Redhat will not help us.  However note than we don't need much work
to issue something significantly better than RedHat: less than a month
work by 4 or 5 volunteers.  Also notice than RPM world is not limited
to RedHat.

-We were supposed to go the Redhat way then Omega told it was Debian,
despite the preceeding policies being not official if we change again
our credibilty will plummet.


Conclusions:

So if we get _real_ help both in quantity and quality then Debian is
perhaps the best way but if we are forced to rely on our own forces my
feel is than we will pay a high price for not trying to start from a
distribution nearer to our goal.

Notice than Bruce Perens was very evasive about Debian help and than
the Debian people will have to learn how to think in end users.

All the project is stacked in the question of Debian help and I fear
what will happen if it does not come or it is weak or unadequate.

In fact perhaps the real interest in cooperating with Debian is
bringing some SEUL spirit to it.  I find bad enough for Linux having
one all-hacker distribution (Slackware) without having a second
(Debian) not caring about end users and beginners.  But that means
than we shall be the givers not the takers of help and don't forget
that for making a user-friendly Debian there is already LaetOS.

Going Debian forces us to a cathedral model of development because
Debian requires extensive changes before we can release something
credible from a user friendliness view point (ie everibody will laugh
if SEUL 0.0.0 is harder than usual distribs).  The problem about
cathedrals is than 50 years after beginning construction you are still
getting wet.  This project has been created in may 97 and nine months
later we are still at the same point.  I do not think we can allow us
to spend another year.

The model who works in free software world is the bazar model: alight
enthousiasm by issuing something credible asap and then make
continuous changes to it.  That means than our starting point must be
the distrib the nearest to our goal we can put our hands on.

Hurd used cathedral, Linux used bazar.  Where are they now?

-- 
			Jean Francois Martinez

==================== The Linux.  Use the Linux, Luke! =======================

===
SEUL-Leaders list, seul-leaders-request@seul.org
===