Open letter to Debian community
Open letter to Debian community.
Dear friends and collegues!
I'm one of the russian Debain community. I'm not famous in the community
in Russia, but I decided to write this letter. I want to tell you my opinion
about current state of the Debian GNU/Linux and discuss some ideas.
I'd like to say. that I don't want to hurt somebody's feelings. I'd want
all interested people to think about questions that I ask here.
So let's start.
I installed Debian 3 years ago. Before this I worked with Slackware. I
remember my first feelings and disadvantages I met, but then I saw
convinience
and beauty of Debian. Till this day I appreciate Debian in terms of
organization of directories, dpkg and number packages it contains. I liked
Debian a lot.
However, today, Debian way of development doesn't go I'd like and want
it to do.
First defect, IMHO is that Debian is very big. I thought and I think now,
that best size of distribution is one CD. One CD for binaries and one for
sources. Some people can say, that I contradict oneself, at once I want
fullness of distribution and little size, but I think that this things can
be combined together.
Today, Potato has many useless packages. I don't say, that they are not
needed by everyone, no, not everyone, but for majority, Who needs baby
shell?
How many IRC clients, ftp clients, and finger, and mail clients! For eg:
how many MTA do we have? Three: sendmail, exim, smail. And IRC clients?
bezerk, bitchx, bitchx-gtk, epic, epic4, irssi, liece, sirc, tinyirc, tirc,
xchat (in three modifications), zircon. 12 pieces! And how many finger
servers? cfingerd, efingerd, ffingerd, fingerd, xfingerd? 5 pieces. etc.
And everything takes space! And games! Do many people play XBill?
And tell me please, why docs about anarchism were included in doc section?
And Bible? Religious people have printed variants and non-religious won't
read it from computer.
And how many old things are included in distribution! Do you think that
is rational to keep Tk7.6 and Tcl7.6 only because of TkDesk? And how many
version of Gtk does distribution have? Three! gtk1.0, gtk1.1, gtk1.2. If
there is software that doesn't build with gtk1.2, then it's place in the
trash not in distribution.
I call leaders to revise principles of forming of the distribution. I
want distribution so contain only needed things and only best things, not
all possible. Distribution needs IRC client. Ok, it will be THE BEST
irc client. Distribution needs games, ok! Let it contain games that are
interesting not only for children from 6-10 years.
I don't like the mess that happens with Potato's package's suffixes
"-dev" and "-doc". It is thought that -dev should be in dev section and
doc should be doc. But in fact they can be everywhere!
Another disadavantage: I analized Potato from the point of russian user,
that doesn't need packages that are not related to russian or english
languages. I found, that I can drop 100Mb packages! I don't need chinese
packages, because I don't speak chinese for eg. You will say, that Debian
is used by many users, not only russian. You're right, but what prevents
to divide Debian on international and national parts? Imagine how easier
it will be? Than for example me, russian, could take international part
and russian part, and japanese people could take international and japanese
parts. In this case european users don't need to know anything about canna,
wnn, chinese fonts, etc, because they simply don't need this packages.
So, we could to have boot floppies and defauls settings for applications
for needed language. In this case, Debian will become Russian, Japanese,
France etc directly! And the size of distribution will be smaller.
Many people will disagree, and say that they don't worry about number
of CDs Debian need. I don't complain, but for me it is hard to afford
more then 2 CDs. But don't think that only because of this reason I agitate
to decrease the size of distribution. This is one of the reason of potato
long
releasing. It is harder to support that big amount of software.
One another defect of Debian are release dates. Potato hit the record
for release date. Delay with Potato was not comparable with other versions
of
Debian. Than Potato released. From one point of view it is good, from
another
we have Debian version that is very old compared with other distributions.
It is not needed to say, that half a year of Information technology, and in
world of Linux is a very long time!
Now I'd like to talk about standards. Standards is a basis that Linux
and Debian need. No, there are standards now. But there are some areas
that don't have standards. For example network services. We have lots of
them: http, ftp, samba, NFS, IRC, ICQ etc. But imagine, that you want to
know what services are currently installed. It turns out, that
the only way to know it is to look what packages are installed. But it is
very inconvenient! It is better to have ONE file that has descriptions
of this services. Make a policy and make network service's packages to
register themselves. WE can go further and think about /etc directory
where information will be stored not only about net services, but
about may other things. For example:
/etc/registry
|-X11/
| |-window-managers information about window managers
| +-xservers information about X servers
|-network/
| |-servers information about network services
| +-clients information about clients for network services
...
I am confident, that you can make this list longer. From my point of
view,
if this worked it made life of the users and programs easier.
In version 1.3 (and in 2.0 AFAIR) there was possibility to turn
off xdm start, even if xdm is installed. It was done in
/etc/X11/Xsession.options by adding keyword no-start-xdm.
However, in late versions this option was been removed. But, sir, if
I installed xdm, it doesn't mean that I want to run it. xdm is only
an example. I want to say, that it would be good to have /etc/daemons
file, and every string was in format
daemonname: yes|no,
and every script from /etc/init.d directory that runs daemon should
see /etc/daemons file. This will make more easy daemon mangement.
This idea is not new. One friend told me that HP/UX has some file
like this.
And in the end I'd like to say some nasty words about dpkg.
Yes, it is very cool. But don't you think that during existance
and development of Debian dpkg developers didn't add new things in
it. Isn't it stagnation? Some "features" of dpkg make me cry.
Don't you think, that names libxxx0-1.deb, libxxx1g-1.deb, libxxx1-1.deb
are very strange? What is it? It is one library, but of various versions.
And this happens because dpkg cannot process existance of two various
versions of came package. As a result, we need this names to make no
"upgrade" and "downgrade". Information of versions that can co-exist with
this package can be added to control file. I have another several ideas
about dpkg, but I'd like dpkg to get rid of this first.
I'm not quite sure, that people and especially leaders of Debian will
percive this letter seriously, but I hope so. However, if anybody will
think about my words, I will be glad.
I'm sorry if thoughts and ideas written here are old, not original and
were discussed later. Because of my bad knowledge of english, I can't
take part in live discussion of all events of Debian community.
I'm sorry for my emotions, but it is because of that grieve for Debian
and I'm upset when I see something bad in Debian.
Translation made by Peter Novodvorsky.
If someone disagrees he can write me directly to my e-mail
victor_v@permonline.ru
Victor Vislobokov
Perm, Russia
Reply to: