Installation Profiles [was: Re: Reality check!]
just as a side note before i proceed: This discussion could probably
easily lead to the impression that i'm considering the already done
work to be really bad, but actually rather the opposite is the case
and i'm thankful that people do what they do. My respect goes to all
those who make the boot floppies - especially after having toyed
around with their setup myself! It's not an easy task...
I'm really not intenting to step on anybody's toes with my critique
but sincerely hope to provide another perspective with my probably
rather awkward usage of the english language. So please don't be
annoyed just because my talk is so much cheaper than your's. ;-)
firstname.lastname@example.org (Stephane Bortzmeyer) writes:
> <email@example.com> wrote:
[ boot floppy sources ]
> You can get the last version by CVS:
Wow, this was just the information i needed. Unfortunately i don't
have access to the net from home anymore, but i will check into it the
next time i'm at the institute. Thank you.
[ broken yes/no answering at the configuration stage ]
> This bug has been reported #32324 and fixed since.
Okay, so i just stumbled over it because i used an already outdated
floppy set as found on our german Debian mirror site.
[ redundant emacs versions ]
> Well, I'll suggest that for potato. It will start a nice flame-war on
> debian-devel "emacs vs. xemacs".
Hey, that's just what we need at this stage for *slink*! >:-)
Okay, let's be serious again: unfortunately this actually means that
some of the most obvious installation profiles of slink stay to be
unnecessarily bloated. I consider this to be a bad move because the
initial install is something like Debian's advertisement plate (or
visiting card) and the installation of three emacs variants gives a
rather bad impression IMHO. I mean, who would *really* want to have
*three* emacs variants installed at once and above all right at the
first installation stage?
Myself i do prefer XEmacs over all other variants but wouldn't mind if
i had to install it later on my own. The reason for my critique was
just to find a way to both cut down on install time and disk space for
the first installation.
> > leaving all other variants as an option for later installation to the
> > installer's discretion. Likewise for the vi variations. Which emacs
> > or vi to use is a matter of personal choice of the installer.
> This contradicts the whole idea of profiles. A profile is a predefined set of
> choices that *we* think OK and the installer which chooses a profile trusts us
> blindly. I regard the Average User as unable to choose between emacs and
> xemacs at the beginning (or between exim and sendmail or between apache and
> roxen). So, we choose for him.
Well, this appears to be more like double thinking (although well
meant) to me. On one hand we strive for making a distribution not
explicitly meant for dummies and then create installation profiles
which a thinking human being wouldn't really want to install. :-/
Honestly, who would really want to have three emacs variants installed
when one needs only a single one? Which of the emacs variants to
install is IMHO largely a matter of personal preference and therefore
one *has* to choose for oneself anyway. So why force all three
variants down one's throat instead of providing the least common
denominator only? Regarding size and currentness this would probably
be emacs20, which i don't really like to use myself BTW. I consider
it nicer to replace the default emacs version upon a later stage with
my preferred one than to have to uninstall the two variants i don't
want *after* having suffered the lengthy installation of all three.
Guess what's less time consuming? ;-)
I personally consider this a nuisance and would rather prefer to
choose the "Basic" install option because the defaults of the given
profiles can result in grieve on the first time installer's side.
This is unfortunately *not* obvious at the first time installation and
first time installers *will* therefore have to learn the hard way. No
wonder that Debian still has the reputation of being too hard to
It's one of Debian's strengths IMHO to give freedom of choice but we
really can't (and shouldn't even try to) take away the need from users
to choose by themselves. Let's face it: they will have to choose
sooner or later anyway because this is not MS-Windows or MacOS. I
wouldn't want that an installation routine forces a not so obvious
oversized default down my throat and i severely doubt that the
targeted average user would be capable of deciding whether a given
profile is *really* suitable for him or not. So why not better reduce
the profiles? I mean "less" is often enough "more"! Too bad if this
would already be out of question for slink... :-(
> > I think it is a very bad habit to first fill up the disk with
> > redundant selections and then expect the installer to deinstalll what
> > [s]he doesn't like/want in order to make room for other software.
> This is a typical example of the main problem with the "Let's make everything
> easier for Joe User" approach: nobody agrees on what is easier.
Maybe we should rather decide whether we primarily target Joe User or
not? Even with the most perfect profiles i doubt that Debian would be
a good choice at all for the average newbie. I always thought that
Debian was rather meant for competent thinking people who can be
expected to choose by themselves?
I would really like to see an installation routine suitable for the
average user. But this definitely needs a complete overhaul of the
whole installation routine, which is not feasible at this stage and
would probably require someone very dedicated to ease of use upon
implementing such a thing. As i wrote before this would be probably a
perfect field of undertaking for a commercial company based on
Debian's free work.
> For me, I think that most users expect things to be already there ("I've read
> in an Unix manual about tcpdump and Debian hasn't it. This distribution is
> broken.") without a new installation, which will certainly be painful for the
> typical user.
I was actually approaching this from the other direction ("Debian is
broken because it takes ages filling up my hard disk with stuff i'll
never use anyway!"). I think a Debian user can be expected to be not
a dummy and he will *have* to make a choice anyway. I've heard enough
people critizising Debian for it's "bloated profiles" (not my wording
BTW!). But as it seems it is already too late(?) for slink...
> > machine, but possibly far less capable hardware. The wealth of
> > software coming with Debian doesn't mean that everything and the
> > kitchen sink should be installed.
> Most of the messages I received, as the maintainer of the list of pre-defined
> profiles are "XXX is missing, why don't you add it?".
I think this is a problem of the right choice. One just can't make it
right for everybody and it is no good idea to add things in order to
please everybody (i vaguely remember an article by Alan Cox about the
town council and whatnot in this context). IMHO it would be much
wiser to provide a useful *minimum* in each category upon which people
can base their own choices. As it stands "Basic" is far too basic for
being an acceptable minimum and the other profiles are far beyond
being a truly useable minimum. There has to be something in between
for each category. Throwing in everything and the kitchen sink IMHO
doesn't really scale at all.
> > What i'd like to see is something like "profiles/Textprocessing" for
> > the writing people containing the TeX system and text/PostScript
> > related utilities. In any case i'll try to make up such a selection
> > and send it to you ASAP.
> Be my guest.
Would this still go into slink or into potato?
Well, i guess my critique is not really suitable at this time and as
stated before i really don't want to defeat anybody's already done
hard work. But i definitely think that in regard of the profiles
"less is more" is the most appropriate perspective.
Thank you, P. *8^)
--------- Paul Seelig <firstname.lastname@example.org> -----------
African Music Archive - Institute for Ethnology and Africa Studies
Johannes Gutenberg-University - Forum 6 - 55099 Mainz/Germany
------------------- http://ntama.uni-mainz.de --------------------