On-disk format of UNIX filesystems (Was: Re: make DOS a filesystem?)

[Andrew Hume]

Submitted-by: andrew at alice.att.com (Andrew Hume)

In article <563 at usenix.ORG>, aglew at crhc.uiuc.edu (Andy Glew) writes:
> ... I do rather
> wish that there were some standards for the on-disk format of UNIX
> filesystems.  Or am I the only person that has ever tried to transfer
> UNIX filesystems on floppies between different systems?  Or (soon)
> transfer UNIX filesystems on floptical disks?
> 
> Most of the filesystems standards work seems to be technology specific
> - such as, the soon-to-become-official standard for CD-ROM filesystems
> and other optical disks.  However, what I've seen of the CD-ROM
> standard suggests that I am unlikely ever to be able to mount a CD-ROM
> as the boot partition of my workstation...
>     Q: what is the UNIX community's particpation in various
> technology-oriented filesystems standardization efforts?  Does everyone 
> feel confident that present and future UNIX filesystem semantics will be
> completely supported by these standards?
> 
> --
> Andy Glew, a-glew at uiuc.edu [get ph nameserver from uxc.cso.uiuc.edu:net/qi]


	the x3b11.1 work that i snitch on is aimed precisely at this.
the current work is aimed at WORM technology but the next standard is
for re-writable optical media - which is isomorphic to regular
magnetic disks. however, it is important to note this is an interchange
standard. for various (read performance and economic advantages) reasons,
vendors may always choose a different format for internal use but at least
you should be able to convert this to the interchange format for carrying
the data around. and in x3b11.1's case, we are tryingvery hard to make
the format a plausible one for use as a regular filesystem.
	i agree the CD-ROM standard does not sit comfortably with Unix
but what can you expect when the primary vendors represented on the high
sierra committee were MS-DOS and VMS? x3b11.1 has active participation
from bell labs research (me) and Sun (tom wong) and HP (ed beshore)
to name prominent Unix representatives; in addition, many of the other members
are acutely aware of the importance of the Unix market.

	as for support of present and future unix filesystems, we are
deliberately adding support for arbitrary additional fields per file-like
thing (as extended attributes) so as far as it is possible, we should be able
handle most future extensions. as for present systems, it is up to the
unix community to comment NOW on what fields are necessary. standard
things like BSD/SysV inode fields can be taken for granted but
perhaps you know of others (file effective date? file expiry date?
automatic logging on write access). please mail such suggestions to
andrew at research.att.com. the ball is in the unix community's court.

	andrew

Volume-Number: Volume 21, Number 164