Bug#1187: Various `vi' versions trample on each other.
email@example.com (Ian Jackson) said:
> Please could all `vi' versions:
> (a) always - without prompting or producing large amounts of output -
> install themselves as /usr/bin/vi in the postinst if no vi is already
> installed. If they do this should always install a link
> /usr/man/man1/vi.1 (even if there is already such a link).
I don't like the way this is currently handled (by my elvis package
or by other vi clones). However, what is currently done was arrived
at after some considerable discussion. I implemented it first in elvis,
and asked that others look at it and speak up if they disagreed. Apparently,
now, the processing I put into elvis has been picked up by nvi. I don't
know if there are other vi clones in the distribution or whether and to
what extent they may or may not have picked this up. Let's make sure that,
if we're going to change our minds on this again, we're sure that we're
going to stick with whatever we change them to this time.
As I understand it, you're asking that the processing be changed so
that programs would install themselves under package-specific names,
and would optionally link package-specific binary names to more
generic names ("vi", "ex", ...) if those names are not in use.
The first vi clone installed, then, would use whatever generic names
it wanted. vi clones installing later would only use names which they
wanted and which were still free (perhaps "ex", but not "vi").
You're also asking that package-specific manpage names be used, and
only linked to generic names if the binary link is done, and that the
manpage link be put in place of whatever file or link might pre-exist
If whatever package had put the overwritten manpage file or link in
place, then, is de-installed; the manpage link will disappear.
Is this the desired behavior? Are we all agreed on this once and for all?
Will someone take responsibility for documenting this in a maintained
packaging Guidelines document so that there's at least some slight chance
that maintainers adding vi clones in the future will follow these special
guidelines for vi-clone packages?
[ObMention of package-granularity vs. file-granularity handling of
dependencies and conflicts]