[prev in list] [next in list] [prev in thread] [next in thread] 

List:       freebsd-hackers
Subject:    Re: Union mounts and other mounts
From:       Terry Lambert <terry () lambert ! org>
Date:       1996-07-30 21:48:55
[Download RAW message or body]

> I understand that Jeff's Lite2 work needs to be incorporated before
> any fs commits can be made.

Yes.  They are the baseline.

Kirk's work is somewhat of a monkey wrench, since it will change the
baseline, perhaps significantly: I don't know how he is going about
the soft update integration.  I exchanged some email with Ganger
right after the paper came out, and he seemed reluctant to go for
a generic implementation using graph reduction theory.  I don't
know whether to blame it on graph theory, group theory, and
topology being ...ahem... rather esoteric, or that once you have
an FFS that has it, you shouldn't need any other FS.  8-).


> I think we're a ways off from any fs commits anyway.  There's been very
> little analysis on this mailing list of the direction the file systems
> might be taking. 

Actually, there's been quite a lot of stuff on this, at least for
local storage.  What's been missing is distribution and replication.

The "WebNFS" modification is an easy hack.  As I stated on Usenet,
it's pretty obvious that it's for downloading to embeded systems without
local storage -- ie: set top boxes.

Vendor		| FS		| For download of
----------------+---------------+------------------------
Microsoft	| CIFS (SMB)	| ActiveX (OLE) controls
Sun		| WebNFS (NFS)	| JAVA Applets
----------------+---------------+------------------------

I don't think this is a direction we need to worry about for quite a
while.  More interesting is transactioning, reliability, and multifile
idempotentence (Tuxedo-like capabilities for transaction interdependence).

Other directions include componentization (one FS quota module, etc.),
hosted metadata for interoperability (one hosting module, ala UMSDOSFS),
and event notification to user space processes (ie: "file browser, the
directory whose icons are presently being displayed, has changed") to
get rid of polling and go more towards a user interactive model, etc..

Transient connectedness and replication are already being looked at by
the Japanese "Nomads" group.


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.

[prev in list] [next in list] [prev in thread] [next in thread]