hpa, you are right. if the target file exists, it should be used.
but if not - why not evaluate the symlink ? i think this is much cleaner than
the current name trans design, and i hope it will replace it.
>Anything else would mean redefining an existing, well-established concept,
>which probably would be detrimental to security, as programs would think
>they were doing something else than they really were.
which is sometimes necessary. init has no option to read a different config
file. so if you want to share / with nfsroot to several machine, it would
be nice to have /etc/inittab a symlink to /etc/GENERIC/inittab.${host}.
and if no file with "inittab.${host}" exists, the kernel could look for
"inittab.myhostname", which is what people need.
>N.B.: For /home this is probably done better with autofs anyway.
i agree. please look again at this as a possible replacement for
name translation.
the name translation uses :
/etc/inittab
/etc/inittab#host=myhostname#
to override the inittab ob host myhostname. this requires scanning the
directory, which makes things slow. and so people want to turn name
translation off for most processes. and it requires to run e.g. nfsd without
name trabslation. but with this approach, all this will not be necessary,
as the varlink effect only comes with "broken symlinks".
all normal operation will as fast as without.
nfsd doesn't care about it - the symlink resolution is done on the client
side, not on the server
tar can backup everything (but it has problems with nametrans).
please give it a chance.
andreas
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu