RE: [Bug 874] lsbinstall should not include -t package

 
>[email protected] changed:
>
>           What    |Removed                     |Added
>---------------------------------------------------------------
>-------------
>             Status|NEW                         |RESOLVED
>         Resolution|                            |FIXED

Similar to comments I made on bug 897, ignoring this
issue does not make it go away.

Recently listened to a keynote by Andrew Morton in which
he observed that Linux Torvalds is fond of saying his
biggest value to the kernel community is in rejecting
patches (presumably because they're wrong, inelegant,
whatever), but Andrew disagreed with this because the
underlying problem is still present, but now a solution
has been discouraged: the kernel didn't get better
because the patch is rejected, something's still wrong.

For the LSB, if adding something that's been requested
several times is the wrong way to solve the problem,
then we should say what the right way is.

So I propose we create a documentation bug category
(we could use the LSB book, I suppose, if indeed we
update the on-line copy independent of intended pub.
schedules of the printed copy) and when closing issues
like this one, open a new bug there (or reassign
bugs to that category, or whatever makes sense).

As with 897, some guidance for LSB developers is needed
to mitigate the issue.  If indeed we can make a case
that a packaging approach that needs a way to get at the
underlying package installer is wrong, and the correct
way to build a package which must first do some interactive
work (at least rpm is probibited from having any
interactive interaction, so typically packagers would
choose to wrap the install in a script) is "....".
If we can't find a way to cleanly describe how to avoid
the need that led to this feature being requested,
then we've got a problem.