Re: NFS -- Hurd
Kalle Olavi Niemitalo <tosi@ees2.oulu.fi> writes:
> tb@MIT.EDU (Thomas Bushnell, BSG) writes:
>
> > Heh. My best answer is that NFS will only support passive translators
> > by extending the protocol, and then it will only do anything if the
> > server supports it.
>
> Assume this was some Hurd-specific protocol. Where would passive
> translators then be saved?
How does the following sound:
NFS does not allow passive translators. They are not stored; the call
fails. If the user wants to set up passive translators, e can use
shadowfs to slap a writable filesystem over top. If the user is using
a diskless machine, the startup scripts can create the correct passive
translatos on a RAM filesystem, say from a tar file.
Alternatively, when a passive translator is "saved", what is actually
stored is simply a command line, no? Perhaps one could simply store
that in some file on the NFS server (possibly even some automagically
named file). One could possibly even arrange
it so that a non-Hurd-extended server could cope with these files.
> When the client sets the passive translator of a file to
> "/hurd/ext2fs /dev/hd0s2", the server must not attempt to
> activate the translator when the file is accessed. And if the
> translator is set in the server, the client must not run it.
> Translators set by the client and the server must somehow be
> saved differently. Is there a Right Way to do this?
Well, what is actually done when this translator is set (as I
understand it) is to write into the ext2 filesystem the string
"/hurd/ext2fs /dev/hd0s2" with some special flags. If one translates
this into some special file on the NFS server (say, every directory
has a file ".hurd-passive-translators") containing the same
information, you get the correct behaviour.
If the server happens to be a Hurd machine as well, the NFS server
process may stumble across some passive translators in the underlying
filesystem. If this happens, they just look like rather odd regular
files. The client can't manipulate this part of the server's
filesystem. Perhaps it should be able to; is this what you're asking?
That *would* require some Hurd-specific NFS traffic and all sorts of
wierdness would ensue (like, say, Linux machines being able to set
passive translators on Hurd NFS servers).
> I fear this is a situation in which Unix shines by simply not
> having passive translators.
I'm not sure you can call it shining when you can't shoot yourself in
the foot because the gun has no bullets.
Andrew Archibald, busily bludgeoning himself to death.
Reply to: