On Wed, 30 Jun 1999 20:27:53 +0000 (/etc/localtime), Hans Reiser
<reiser@ceic.com> said:
> So, "dirname/..body" opens dirname like it is a file, and the user space
> library translates "dirname" to "dirname/..body" when it notices that the
> underlying FS does not support overloading.
That's exactly the sort of thing I had in mind, yes. If the client NFS
software simply cannot deal with a cached file being part of a filename
component, then yes, we do this. If we can seriously address the
portability problems by offering concrete proposals for getting round
them, then we can really start to persuade people to migrate. If there
is no interoperability between the old and new systems, then that
migration is a lot harder.
Note that in theory we can even do the same form of namespace munging on
local filesystems, so that people can run the extended semantics on
their ext2, vfat or iso9660 filesystems by adding a user library or
wrapfs stacked-filesystem layer on top. However, if the user happens to
be using a filesystem which understands the semantics natively, then we
get all the performance advantages of avoiding the extra namespace
lookups.
--Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/