Re: Switching the debian-edu build rule to use cdd-dev?

[Petter Reinholdtsen <pere@hungry.com>]

[Andreas Tille]
>> - the tasksel tasks are rewritten to use one file instead of several.
> 
> Is there any reason that could convince me to follow this change
> in cdd-dev?

Reduced complexity is the main reason I did this.  Instead of having a
base file, a script and several task files, it is enough to have one
file only.  Also, it reduces the name space "polution" as only one
file name is used in the tasksel directory.

> I guess this is related to Debian policy.  Are non-free packages
> allowed as recommends?

I believe it depends if it is the only package listed in the
dependency set or if it is one option among several where at least one
is in debian/main.  I believe "Recommends: main-pkg | non-free-pkg" is
ok, while "Recommends: non-free-pkg" is not ok.  I am aware there are
some people claiming that non-free packages can not be listed in
depends or recommends at all because that would be to advertice for
non-free software.  I do not share that view.

> Any idea whether we should also include Enhances?

Perhaps.  Is it implemented by apt(itude)?

I forgot one change I did to the package sorting order, affecting the
CD build.  It used to sort packages "task1-depends, task1-recommends,
task2-depends, task2-recommends", and I changed this to
"task1-depends, task2-depends, task1-recommends, task2-recommends", to
make sure the highest priority packages had the highest chance of
making it into CD 1.

> I'm willing to make the move as easy as possible.  On the other hand
> I'm missing kind of a "list of requirements" from Debian-Edu.

Yeah.  I guess we should try to write such list.  At the moment, I
could boil it down to "generate the same package list as
gen_control.pl", thought I guess you want it slightly more
detailed. :)

> Sure.  I guess you will only find a reason if cdd-dev supplies so
> interesting features that it is attractive enough to leave "the old
> path".  I'm thinking about supporting DebTags but that's currently
> vaporware.  I just want to do some research how good are the chances
> to convince other projects to switch ...

Yes, finding some advantage for cdd-dev would increase the priority of
switching. :)

Friendly,
-- 
Petter Reinholdtsen