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

dpkg v2 ? I haven't heard about it, but ...



After reading a message on -devel-announce, it's obvious that I have
to take more seriously a rumour that I heard: namely, that a bunch of
C++ weenies[1] wanted to rewrite dpkg.  There are a number of things
I'd like to say, in no particular order.

* I don't think that most of dpkg needs to be rewritten.

* I am actively working on dpkg.

* Many of the problems that are in current dpkg versions (particularly
the build system) are the result of NMUs.

* I shall be releasing a new maintainer-upload of dpkg into unstable
Real Soon Now (tm).

* One of the main reasons that dpkg is unapproachable by poor to
mediocre programmers is that it must performs a complex task with a
very high level of reliability [2].

* I am distinctly doubtful that the core functions of dpkg are likely
to be rewritten competently in the near future.

* A number of competent people seem to have little difficulty working
on dpkg.

* I agree that there are changes needed.  Of the most urgent changes
are:

 - Reengineering of dependency checking so that it happens in one
   place.

 - Programmatic API for UI developers.

If you wish to reply, please do so on debian-dpkg.  The volume in
debian-devel prohibits reading it.

Ian.

[1] Please do take offence if you're one of those people who think
that C++ and object-orientation are the right solution to nearly every
serious programming problem, or if you find most of the dpkg source
code difficult to understand.  Otherwise this insult is not directed
at you.

[2] The kind of reliability that it is required is that dpkg does not
break systems.  Incorrect error messages, crashes, etc. are of course
not good either and dpkg shouldn't have them, but not breaking
systems, even (especially!) under error conditions, is very important.


Reply to: