Andreas Hauser wrote:
When labor is not that cheap, you need better technology
to accomplish similar. That is what ports/pkgsrc is. It makes
producing those packages much easier, so that less people
can produce more packages.
It is not so difficult to produce Debian packages. I have played
a lot that game to inject more recent software in woody. If the
software has a configure script, which becomes common, running
dh-make on the source code produces the debian subdirectory with
debian/rules, and usually very little hacking is necessary to produce
a working .deb.
I am sure one could easily modify ports/pkgsrc to produce dpkg or rpm
binary packages instead of pkg packages. But this is an ortogonal
question.
I think, what you really appreciate, is that debian has a nicer
way of _distributing_ the packages (and the info about their state).
One could achive similar by freezing a ports tree and producing
a set of packages. Then distribute a ports tree lite [1](only the
version and dependanycy info). And use portupgrade -PP.
I think you have an important point. The only way to have a reliable
packaging system is to produce binaries of *all* the software you
claim you are releasing. With source based systems there is always some
software with doesn't build at time t, and jeopardizes building of all
dependant software. This happened to me countless times with Gnome stuff
on FreeBSD. Now this presents problems, since some software apparently
cannot be delivered in binary form, and has many dependencies, the worst
offender being Java. I don't know the solution to this problem, but it
is probably the main obstacle to a really working portupgrade.