perl and perl-modules; reflexive dependencies vs. archive bloat
Currently there is a proposal to combine perl-modules and perl into
a single package. perl-modules currently contains architecture
independent bits, and perl contains architecture dependent bits.
FWICT, the primary argument to do this is because perl and
perl-modules both require the other to function properly (so a
reflexive dependency or combining is reqeuired), and because circular
dependencies make things complicated for dpkg and other frontends.
Because this is a common situtation, where there is architecture
independent data (of varying sizes) which is interdependent on
architecture specific code of a particular version, reflexive
dependencies are necessary.
Are there still tools which are in use which are unable to handle
reflexive dependencies of this type? If so, does the effort to fix
them outweigh the effort to remove all of the circular dependencies
and the resultant archive bloat?x
0: http://lists.debian.org/debian-release/2009/10/msg00226.html (cc'ed
to -release, but further discussion should not happen there because
release isn't a discussion list.)
2: The alternative is of course, archive bloat, where we have multiple
copies of such data; the instant case will only contribute around 150M
(3.1M*(13 archs-1)*4 releases) to the size of the archive, but I've no
doubt that there are larger examples.
3: Specifically, where Package A Depends on (B=1), and Package B
Depends on A; A and B are from the same source, B is architecture
independent, and does not require configuration.
The computer allows you to make mistakes faster than any other
invention, with the possible exception of handguns and tequila
-- Mitch Ratcliffe