[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: "dselect" replacement team

I've resubscribed to debian-devel, after letting things slide (was
auto-unsubscribed some time back, I don't know why).

I'm fairly dismayed by some of the bickering I see on this dpkg
rewrite issue.  There's some real issues here, but I dread looking
forward to day after day of this stuff.

I'm wondering if maybe we couldn't formalize the process a little bit?
[You know:  project proposal, requirements analysis, design,
implementation, testing, release, recurse.]  Maybe this is already
happening, but I can't figure out what requirements are being
discussed, in many of the messages.  And I'm fairly certain that the
choice as to what to keep/scrap should directly follow from the
requirements of the intended new release.

Note that for bootstrapping purposes we could engineer a "dpkg-lite"
that leaves out a lot of the safety checks and so on and just unpacks
and configures (and maybe a dab of topological sorting).  You don't
have to worry about keeping an existing system running when brining up
a new system, and maybe it's a bad idea to constrain the full-fledged
manager to the requirements of bootstrapping.  [Note: forking
development like this puts greater demands on testers and the testing

Aside: if you want to blame the current defects of dpkg on someone,
you might as well blame them on me, rather than Ian (I was the one
originally tasked with documenting the system -- and I've dropped the
ball on that more than once, and in a major way).  I never have fully
understood all the ramifications of dpkg, and I'm glad to see someone
with more energy than I willing to tackle the project.

To anyone who is tempted to say: "don't do that", Please, try to
re-cast your idea as a test-suite concept (or whatever).  There's
innumerable issues here, and tackling them for real is going to break
some things.  The trick will be to (a) understand what the issues are
and (b) adequately testing the system.

[I hope people take this message as technically challenging, rather
than personally challenging.  I'm very sorry if I've offended.]


Reply to: