NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/protocol/nfs/2953

Text File | 1992-12-14 | 2.7 KB | 66 lines

Newsgroups: comp.protocols.nfs
Path: sparky!uunet!eco.twg.com!twg.com!twg.com!hyung
From: [e-mail] (Henry Yung (The Last In Line))
Subject: Re: NFS cookies
Message-ID: <[e-mail]>
Lines: 53
Sender: [e-mail] (Henry Yung (The Last In Line))
Reply-To: [e-mail]
Organization: The Wollongong Group, West Coast Operations
References: <[e-mail]> <[e-mail]>
Date: Mon, 14 Dec 1992 22:20:36 GMT


In article <[e-mail]>, [e-mail] (Curt
Finch 903 2F021 [e-mail] 512-838-2806) writes:
|>
|>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.
|>

According to the protocol, that is true, but unfortunately, in
reality, that is not always the case.

|>I vaguely remember seeing code somewhere in someone's NFS implemen-
|>tation, (And I've seen lots of implementations now,) where the client
|>used it to see how far to increment some value for the next readdir (or
|>something like that.) If the client's doing that, he's broken.  He's
|>violating the protocol.
|>

Unfortunately, the cookie is not THAT opaque.  A lot of clients do
care about what is inside the cookie, including Sun's.  A couple of months
ago I posted a similar problem to this group describing a problem where
a 'ls' from a Sun client kept on listing the same set of files again
and again.  It turned out that although my cookies were increasing from 0
onward, the 32-bit word (cookie) was not in network byte order.  The
VAX NFS server, having a different integer representation than that of
Sun's, sent the cookie in its native representation thinking that the
cookie is opaque, which turned out to be not true.  Consequently, the
client was seeing cookies with some huge values and in a strange order.
Changing the byte order of the cookie on the server pleased the client
and resolved the problem.

|>But if some client doesn't work right with your server and their
|>support organization is no help to you, that isn't much consolation is
|>it?

Well, in our case, we changed the server to accomodate these clients.

|>--
|>[e-mail] (Curt L. Finch) | AIX NFS/NIS Field Quality
|>My views are unrelated to my employer's    |        Austin, TX
|> There'll be too many elderly in 30 years for your kids to afford all the FICA.
|>
|>

----------------------------------------------------------------------
  __  __/  |      /    _____/  Henry Yung    Email: [e-mail]
     /     |     /    /        VMS Software Engineering
    /        /  /    / -- /    The Wollongong Group, Inc.
  _/       _/ _/   ______/     Voice: [phone] FAX: [phone]