[Debtags-devel] archive strata

[Thaddeus H. Black]

--opJtzjQTFsWo+cga
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Looking ahead two years or so, some early
thoughts regarding the stratification of the
archive occur to me.  (The thoughts come from a
draft of a mail to Martin Pitt.  Cut out of the
draft, they are now set down here.  A copy goes
to Martin.)

The exercise of ramifying the archive package by
package tends to persuade one

  (a) that the Priority dpkg control field was
      well conceived, was well defined, and is
      even more useful than it appears at first
      glance to be, and

  (b) that the massive `optional' Priority level
      wants to separate into three distinct
      strata.

This is no new idea, of course; nor is it
inspiration.  The contents of the archive
themselves seem to demand it.  The
stratification appears to be a natural one.
Broadly speaking, the optional Priority level is
seen as stratifiable into

  optional-1: packages of widespread interest,
  together forming a heavy but stable, well
  balanced, well tested body, 100 % of which a
  user might reasonably install.

  optional-2: packages prominent within
  significant sectors of interest.

  optional-3: other optional packages.

Such definitions naturally eventually need more
detail, but the right way to detail them is
probably to stratify the archive then to go back
and formally to describe what we did, rather
than the other way around.  In this way,
experience not speculation gives rise to the
formal definitions, and the definitions become
our helpful servants not our arbitrary masters.

With the archive ramified, I would estimate the
initial stratification of the archive to require
about 150 hours of work, to be done by one, two
or three interested people in close
communication.

If I had to guess the fraction of all main
binary packages in each stratum, I might guess
(this is just a guess)

   1 % standard, important and required
  12 % optional-1
  17 % optional-2
  56 % optional-3
  14 % extra

Likely candidates for optional-1 may include the
principal parts of Vim, Emacs, TeX, LaTeX, X,
Gnome, Apache, Ghostscript, Docbook, Ruby, G77,
Guile, Octave, Gnuplot, WordNet, Zlib, Libglib,
GTK, QT, Tcl, Jade and others, plus appropriate
documentation.

We have plenty to do already, of course, and no
one is recommending that we start stratifying
the archive today.  The foregoing are just some
thoughts you might find interesting, a sort of
vague pre-proposal for future activity.

--=20
Thaddeus H. Black
P.O. Box 1020, Christiansburg, Virginia 24068
USA; +1 540 961 0920, t@b-tk.org

--opJtzjQTFsWo+cga
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iEYEARECAAYFAkEFIqoACgkQh3E0gzgBXn4UIwCeKUHIrXFPuiGVdaoi+3Td0TkR
ic0AnRhkGhrUHANfC3UwV4ZkLnCuBqGT
=HQjK
-----END PGP SIGNATURE-----

--opJtzjQTFsWo+cga--