[Debtags-devel] where to end debram development

Thaddeus H. Black t@b-tk.org
Sat, 16 Oct 2004 16:53:55 +0000


--LyciRD1jyfeSSjG0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Enrico's advice is requested---and the same
question is open also to others who have counsel
to add.  The question is, since debram will
cover only sarge, never sarge+1, how much
further effort should go into perfecting sarge
debram?

Recently, I have manually been checking the
ramification line by line.  The check is going
well and fairly swiftly (much faster than the
original ramification effort went).  The check
is discovering and correcting a moderate number
of errors: perhaps three misramifications per
hundred packages, which is about what we would
have expected.  So far, so good.  I have
finished about 60 percent of the check.

Here are my estimates.

  (a) When the present check is complete,
  debram will be about 97 or 98 percent
  accurate.

  (b) It will take me perhaps 70 to 100 hours
  to split overburdened branches up and
  otherwise to reform the ramification as
  discussed recently on this list.

  (c) With 200 to 300 additional hours of
  development effort, we could improve
  debram's accuracy to 99 percent.

  (d) With 300 to 400 further additional hours
  of development effort, we could improve
  debram's accuracy to 99.5 percent.

  (e) With yet 300 to 500 further additional
  hours of development effort, we could
  improve debram's accuracy to 99.7 percent.

The further efforts in (c), (d) and (e) would
include not only careful manual checking and
rechecking, but also the development of
increasingly sophisticated check scripts to
trace through webs of dependency and to
coordinate and analyze other misramification
clues.  Naturally I could not go all the way
through (e) on my own.  If you felt that such
large efforts were better devoted to debtags
proper than spent on an obsolete debram system,
then I would tend to agree with you.

What we would therefore like to know is, would
autodebtags find 97 to 98 percent ramification
accuracy acceptable?  If it would, then I would
like to complete only (a) and (b)---leaving (c),
(d) and (e) undone---after (b), joining the more
interesting work on debtags proper.  Debtags
would not thus be limited to 97 or 98 percent
accuracy, of course; but the large body of data
sourced from debram would harbor perhaps two or
three errors per hundred packages.  Further
gains in accuracy would be made post-debram
only, directly in debtags proper.

Without good advice, I seem unlikely to strike
the right debram balance.  Please advise.  Also
please feel free to pose additional options or
otherwise to expand your remarks beyond the
specific question asked.  Perspective is sought
today.

--=20
Thad

--LyciRD1jyfeSSjG0
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

iEYEARECAAYFAkFxUiMACgkQh3E0gzgBXn4a3QCeMp37q+hwHEJ9GMng5uFQ9ydm
XlYAoMzVHtAg4WoCJ5z7eYHxEnLgYh6F
=1+Es
-----END PGP SIGNATURE-----

--LyciRD1jyfeSSjG0--