graph theory oportunitties in Debian release process
Hallo,
I use Debian and I wondered, why is everything here so obsolete.
Then I wondered, why it takes so long for packages to pass, even to
the testing release.
So I read about process that passes packages from unstable to
testing and then to stable release. I see major slowdown reason in
unsatisfied cross- and core packages dependencies. Often there are few
core packages (glibc, libc6 etc) that block many others. And often some
unbreakable dependency problems.
I feel that today's script based mechanism is too simple in roots
and cannot handle complicated dependency problems. I feel, that some
problems could be unsolvable too. The problems persist and major
releases are too rare and they contain very old software even in time
they appear, not to mention after year, when still no next release is to
see, and everything in system is soooo old. And only the stable release
is the one I can install without basic functionality fear.
I'm sorry I'm not math genius nor programmer to help, but I can
suggest something:
some engine based on graph-theory should be build that will search for
best way that packages should follow to decrease the locks caused by
unbreakable cross-dependencies. It should find the core problem packages
that should be forced forward to allow many others to pass, and find the
shortest path to accomplish this. The engine should take some "value of
importance of the package", the obsolecy (how old is the package in last
stable distribution), take in care more versions of packages (with their
different dependencies), even future versions, and mix it all down.
Output will be the mix of packages that can go thru right now and
packages that should be forced (namely libraries) to allow break the
dependency problems. The goal is to minimalise number of dependency
problems.
What do You think?
Sincerely Peter
Reply to: