NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/protocol/nfs/2943

Internet Message Format | 1992-12-13 | 2.8 KB

Path: sparky!uunet!elroy.jpl.nasa.gov!usc!wupost!cs.utexas.edu!sun-barr!west.West.Sun.COM!cronkite.Central.Sun.COM!sixgun.East.Sun.COM!seven-up.East.Sun.COM!tyger.Eng.Sun.COM!geoff
From: [e-mail] (Geoff Arnold @ Sun BOS - R.H. coast near the top)
Newsgroups: comp.protocols.nfs
Subject: Re: NFS cookies
Date: 14 Dec 1992 00:21:56 GMT
Organization: SunSelect
Lines: 42
Sender: [e-mail]
Message-ID: <[e-mail]>
References: <[e-mail]> <[e-mail]>
NNTP-Posting-Host: tyger.east.sun.com

Quoth [e-mail] (Curt Finch 903 2F021 [e-mail] 512-838-2806) (in <[e-mail]>):
#In article <[e-mail]> [e-mail] (Keith Michaels) writes:
#>Also, is there any truth to the rumor that the first byte of the cookie is
#>reserved for use by the client?  I can't find any code to substatiate
#>that.  If it's true, it also cuts down on the cookie range.
#
#The cookie is generated by the server and is supposed to be opaque
#right?  Clients shouldn't look at it.  They shouldn't even care what's
#in it.

Correct. The only distinguished value is 0, which identifies the first
entry in the directory search sequence. (It would be incorrect for the
client to assume that that's the first entry in the underlying
directory. In fact, the client isn't even allowed to assume that
cookies are ordered in any way...) If the cookie which the client
presents is non-zero, it MUST have been returned by the server in the
reply to an earlier READDIR on the same directory. Thus a client cannot
synthesize a cookie at all.

There are a couple of ways that servers can implement the cookie. The
usual way is to use the offset into the directory itself; this
restricts the number of files to O((2^32)/avg.directory.entrysize). For
a 64 byte entry, this gives you around 67 million files per directory.
The second way would be to use the ordinal of the directory entry,
which presupposes a really efficient indexing scheme..:-)

Actually, there is one other thing that servers can do to try to
improve correctness of the file system model. The server can carve off
a portion of the cookie (8 bits is convenient) and store a version
number derived from the modification time on the directory. This allows
the server to detect a "stale" cookie very quickly, at the expense of
allowing few files per directory. I don't know of any servers that do
this, but in any case such a scheme would be transparent to the
client.

Geoff

--
Geoff Arnold, PC-NFS architect, Sun Select. ([e-mail])
ADMINISTRIVIA==========ADMINISTRIVIA=====ADMINISTRIVIA==========ADMINISTRIVIA
New address: SunSelect, 2 Elizabeth Drive, Chelmsford, MA 01824-4195
New numbers: Phone: [phone]   FAX: [phone]