Re: Moving to the FHS: not right now!
I'm replying to two at once here, in the interest of efficiency.
Chris Waters wrote:
> Yes, that's why I suggest that we wait till after Potato, and start
> the changeover at the *beginning* of a release cycle. That way we
> have as much time as possible.
That was the plan the previous two releases as well :-)
Chris Waters wrote:
> Richard Braakman <dark@xs4all.nl> writes:
>
> > That last sentence is an error. When all packages have moved to
> > /usr/share/doc, we can drop the symlink handling code from the
> > postinst and prerm, with no loss.
>
> Er, no, not if the symlink handling is *in* the postinst/prerm. If
> it's there, then we have to wait another release or two, in order to
> support people doing partial upgrades.
How so? It seems to me that we have to wait exactly one full release.
When we've released a version that has all packages converted, the
next release can drop all the symlinking code.
> > That is because at every upgrade, the symlink is removed by the old
> > package and (possibly) reinstalled by the new package.
>
> Ah, I see. If the symlink handling is all done by dpkg, and NOT by
> postinst/prerm, then you may indeed be correct.
No, it's done by postinst/prerm. dpkg can't do it if the symlink
is in the data.tar.gz, because it will not replace a directory with
a symlink.
> > The only remaining technical objection I have to it is that it will
> > fail for packages that have extra files in their /usr/doc directory,
> > either due to package cruft or because the system admin put something
> > there. Those packages can not make the symlink because there's still
> > a real directory, and it will appear to the user that the
> > documentation is simply missing.
>
> Won't the files just disappear? The way the contents of /dev
> disappeared recently? :-)
I'm not sure, I'm not about to experiment :)
[philosophical discussion deleted]
> But, if some of us think that we're *too* focused on releases, and
> some of us (like me) think that we're not focused *enough* on
> releases, then maybe we're actually striking exactly the right
> balance. I dunno. :-)
I don't know about a balance, but certainly we should have people
watching both sides of the scale :)
Richard Braakman
Reply to: