Re: Why df does not work

[Paul Emsley <paule@chem.gla.ac.uk>]

>>>>> "Marcus" == Marcus Brinkmann <Marcus.Brinkmann@ruhr-uni-bochum.de> writes:

    Marcus> The Hurd is a multiserver OS, and there is no central
    Marcus> authority for mounted filesystems.  What we could do is to
    Marcus> write a database server where filesystems can register
    Marcus> themselves if they want.

	That sound like fun.  How does it work?  This is the server
      which df would used when invoked with no filesystem as an
      argument?  Is there a program which would do something similar
      for some other service?  What should I have read before asking
      you these questions?

	Thinking aloud:

	Let's add an optional call in mount (is mount the right
      place?) a call to
	register_mount_point(host, remote_host, remote_mnt_pt, local_mnt_pt)
        (or some such).

	I guess that register_mount_point() gets defined in a hurd
      library somewhere.  What does it do?  As I understand it, alice
      and bob (as well as root) can be running df servers too. How
      does register_mount_point know to register with those too?

	How does the multicomputeriness of the Hurd fit in?  Each
      computer should have a df server?  Or there is one central df
      server to which df requests (register_me!, what_ya_got?) are
      made from our Hurd cluster?

	What sort of database would be appropriate for a df server?
      A PostgresSQL interface or just a list of lists (that Guile
      could handle).  Would Guild be an appropriate language with
      which to implement a df server?  It seems lightweight, so
      perhaps the answer is "no"...

	Yours ignorantly,

	Paul.


	

--
Paul.Emsley@chem.gla.ac.uk   
 http://www.chem.gla.ac.uk/~paule