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

Re: Relative and absolute symlinks



On Fri, Aug 15, 2008 at 11:42:23AM -0700, Russ Allbery wrote:
> Gerfried Fuchs <rhonda@deb.at> writes:
>> * Lionel Elie Mamane <lionel@mamane.lu> [2008-08-15 16:47:41 CEST]:

>>> So, is there any reason at all to use relative symlinks?

>>  Quite some times I experienced them to be more pain than gain, too. It
>> might be useful if people shift around complete hierarchies, but we are
>> not really speaking of package-internal symlinks here usually.

> And Debian doesn't support relocatable packages in general anyway.

Well, users are not feeling they are "relocating" a package when
replacing one of its directories by a symlink to elsewhere; after
all, paths of the form /original/package/file/patch still work.

> We should clearly use relative symlinks within the same directory, and
> probably from a directory to a subdirectory, but I do wonder about the
> merits of any symlink containing ../.  I'm not sure what we'd lose by
> making any symlink that climbs directories absolute instead of relative,
> and I think we'd definitely gain from having somewhat less weirdness and
> breakage in corner cases.

Well, these symlink for example:

/usr/share/doc/bash/completion-contrib -> ../bash-completion/contrib
/usr/share/doc/bash-doc/examples -> ../bash/examples
/usr/share/doc/cpp/README.Bugs -> ../gcc-4.2/README.Bugs

should IMHO stay relative, because of the scenario described by Manoj,
and users are extremely unlikely to replace a subdirectory of
/usr/share/doc by a symlink (except /usr/share/doc/texmf?). But maybe
managing such exceptions would be too hairy altogether and we'd want a
much simpler policy in exchange for only 99.9% correctness.

-- 
Lionel


Reply to: