unix file system

[Guy Harris]

> The Xerox Star file servers have a file structure which allows some
> ridiculous amount of attribute data (64K?).  As you might guess, attributes
> in that OS are used to hold all sorts of things, including file names!

The nice thing about the Pilot OS that the Star application software ran
under is that the lowest level of the file system gave you the ability to
refer to a file by its unique ID.  It had no notion of file names
whatsoever, and only supported four file attributes:

	Pilot recognizes only four attributes: size, type, permanence, and
	immutability.

(From Redell, D. D.; Dalal, Y. K.; Horsley, T. R; Lauer, H. C.; Lynch, W.
C.; McJones, P. R.; Murray, H. G.; Purcell, S. C., "Pilot: An Operating
System for a Personal Computer", CACM 23(2): 81-92; Feb. 1980 - a good paper
to read if you're interested in operating systems.  Also read the
retrospective "well, here's what we did right and here's what we did wrong"
paper, Lauer, H. C. "Observations on the Development of an Operating
System", Proceedings of the 8th Symposium on Operating Systems, in ACM
Operating Systems Review 15(5): 30-36; Dec. 1981.)

Anything more was up to the next level up.  I'm loath to say "the
application" - Pilot runs on machines with no "protected mode", limited
memory protection (write-protected code pages only), and one address space
for all processes, so it's not entirely clear what's "operating system" and
what's "application".

One advantage of storing a file name with the file "attributes" is that if
the file gets lost the file system salvager can give it a better name that
its unique ID in BCD when reconnecting it to a directory.  (Files-11,
although its basic structure is essentially the same as UNIX - i-list,
called "index file", and directories which merely map names into indices -
stores the file name in the "inode", and can use this when restoring the
file into "lost+found", called [2,3] or something equally mnemonic.  Of
course, if two files with the same name are lost, so are you...)  I only
used a Star once; if it's directory structure is anything like what I infer
the Lisa's is, it's not a problem.  My model for both systems is that you
have "folders" which are directories, but since files are opened by unique
ID (my guess for the Lisa; that's how I would have done it, anyway) they
don't have to have unique names within a directory.  The names are merely
for human convenience.  You can open a file by pointing at it and hitting
the appropriate keyboard/mouse button(s); the system knows the unique ID of
the file whose icon the mouse pointer is in, so again it doesn't need the
name.

	Guy Harris