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

Re: /usr/share/man



On Tue, 21 Dec 1999 Hugo.van.der.Kooij@caiw.nl wrote:


> On Tue, 21 Dec 1999, Julie wrote:
> 
> > From: Rob Lembree <lembree@sgi.com>
> > > >/usr/share/man alone is not sufficient. Not all man pages are platform
> > > >independant. I'd favour keeping both therefore
> > > 
> > > Or even system independent.  Software installed here may not be 
> > > installed there.  I would also vote to maintain both, with clear
> > > guidelines as to why one might choose one over the other.
> > 
> > How is this going to solve the issue of system-independent
> > commands which expect to find system-idependent manpages
> > in /usr/man?  My vote is a symbolic link from /usr/man to
> > /usr/share/man.
> > 
> > I thought the "local commands manpages" convention was
> > /usr/local/man anyway?
> 
> My 0.02 are to go to /usr/share/man and keep /usr/man as a symlink. As far
> as I know it should work for all legacy stuff around but still clearly
> mark the path to go.
> 
> It is known that manpages could be in several places so software should
> handle it more or less dynamically anyway:
> 
> 	/usr/man		traditionally
> 
> 	/usr/share/man		newer
> 
> 	/usr/local/man		for some local software
> 
> 	/opt/.../man		some packages put these in their own tree
> 
> 	/var/opt/.../man	or under the var stuff for packages
> 
> 	$HOME/man		my own kitchensink stuff (manpage make a
> 				great cookbook in you kitchen if you
> 				happen to have a terminal installed
> 				there.)
> 
> I have seen them all except the /var/opt/.../man types.
> 
> Hugo.
I would like to point out something on the system management point of
view, Allan Cox was complitely right saying that /usr/share/man is not
enought, /usr/man should be kept. I agree with all his reasons, but I am
not thinking to a link.

I would prefer to put in /usr/share/man just that man pages
that could really be shared and are not machine dependant, and
in /usr/man the system man pages machine dependant.

I do know that this criterium is really difficoult to define, because
on some point of view also a simple command like "mount" could be
considered machine dependant, but, to be onest, i would not expect
mount(2) or even mount(8) to be in /usr/share/man, 
but i would expect ls(1) or printf(3).

Then there are a lot of other man repository in unix system
(like /usr/local/man), also
depending on the architecture and if the man pages are compressed,
formatted or not. Just think to /var/man/cat[1-n] in a lot
of Linux distributions, to which are linked all the cat[1-n]
directories in all man repository in the FS.

As you know there are packages that have teir own tree, as teTeX, 
and teir own man page repository ${TETEX_DIR}/man/man[1-5].

Too aften system administering get difficoulties from this
disorder, and the ${MANPATH} becomes very long in /etc/profile 
and so on.  

It is necessary to improve the order, so i agree with the 
/usr/share/man criterium, but then we should make
a well pondered distinction about what to put inside

                                  Luigi Genoni
 


Reply to: