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

Re: Packaging and installation

Is nobody else astonished that something so fundamental a standards issue
as packaging formats, that was first debated the moment the LSB started
more than two and a half years ago, is still being argued? This group has
made so little progress in that area that a call to practically start from
scratch is still taken seriously.

The problem is not that the RPM approach doesn't (or, more correctly,
can't be made to) work. It's that this group has never reached a formal
consensus/vote/decree/whatever on a packaging architecture. Had it done so
ages ago (as Bodo suggests), resources could have been allocated to first
draw up an RPM spec that would have at least had all the RPM-using distros
on the same page. Debian programmers would have had an unambiguous target
to parse. Efforts such as LANANA could have been accepted as a formal part
of the LSB; having done that, this group could then address RPM's
deficiencies at a later date.

This step still can and needs to be done. Otherwise, the LSB will still be
arguing packaging formats two years from now, while the rest of the world
moves on without it. How relevant does the LSB strive to be?

Nick's core complaint about the yet-unformalized status quo was:

> if you install anything that a program needs via any other method
> besides RPM, then RPM will consider dependencies unresolved...

Well, duh. If you make something a standard and users break the standard,
then dumb results will follow unless the users know what they're doing.
Creating a standard designed to anticipate how people will break it seems
hardly worth the bother, especially if it means dragging out the decision
to actually *have* a standard even longer.

Allowing for a manual way by which admins can manipulate the RPM database,
to allow a scriptable way to announce the existence of capabilities
installed outside of the RPM system, may be a reasonable compromise. This
is workable so long as the RPM standard itself is well defined.

In the meantime, how much longer must this core issue remain unresolved?

- Evan

Reply to: