> -----Message d'origine-----
> De: Rich McClellan [mailto:[EMAIL PROTECTED]]

> Is it possible to have autofs show a listing of mount points 
> even if they
> have not been mounted? I realize that it might tax the automounter to
> mount every point when someone does an `ls /net` or `ls 
> /home`, but the
> users here are going nuts trying to find there way around blindly.

as told here, today you mut rely on symlinks or similar hacks...

just an idea for a future relase:

why not allowing automoint to answer to  directory listing
by calling some adequate function...

with static file map, it would be just a list of the
fixed keys (ignoring wildcards)
same with YP/NIS+ maps

with program maps it would call a command that should return the list
of supported keys, if some can be guessed (browsing network or servers
exports)...
 and so on...

the duration of this listing should be configurable (cached with timeout or
each time)
and it should still be possible to automount directories that are
not in the list as usual.

I see a problem (but I'm sure since I don't know inodes secrets),
that probably make that the "mount tree" patch is forced
to mount every mount point:
it may not be possible to list a directory without
creating each inode inside, so without mountig each sub mount...

anyway, maybe is my objection false. as far as I know
when you list a directiry you only returns a list of names,
not a list of inodes.
however most /bin/ls options ask for the inode properties
(readable, directory...) which force inode existence...

however is it possible to create empty directoy inodes on which
you can (auto)mount real directories when needed...

it may demand a strange version of inodes
that trigger the automounter only when
someone ask for it's content (chdir, subdir usage or listing)
but not when just asking for inode properties...
the hardest point implementing this
may be to support automount on chdir, while 
not automounting on inode properties...

there should be many inode wizard here to comment my
idea? it it realistic ?

--
Alain Coetmeur, Informatique-CDC DTA
mailto:[EMAIL PROTECTED]

Reply via email to