Date: Tue, 1 Jul 1997 09:28:10 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: tom@sdf.com (Tom Samplonius) Cc: terry@lambert.org, chuckr@glue.umd.edu, lmcsato@lmc.ericsson.se, freebsd-hackers@FreeBSD.ORG Subject: Re: NFS V3 is it stable? Message-ID: <199707011628.JAA18500@phaeton.artisoft.com> In-Reply-To: <Pine.BSF.3.95q.970630194523.16592A-100000@misery.sdf.com> from "Tom Samplonius" at Jun 30, 97 07:50:58 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > > I've had sites report that even going Solaris to Solaris using the > > > Solaris propietary lockd stuff, they still can get trashed mailboxes. > > > Locking of any kind over NFS just isn't reliable. > > > > This is only true if your NFS violates the NFS write guarantees, > > which SVR4 and Solaris do, by default. Turn off write gathering > > and write caching (the "adb" commands are listed in the Network > > Administration Manual) and the problem will go away. > > So you make a trade reliable locking for performance? I'd take > performance any day, and just forget about exporting mailboxes. No, you make a trade: reliable data commits for any software which requires them (CVS, databases, etc.) for "performance". Note the quotation marks: I can make any software run as fast as you want me to make it run, so long as it doesn't have to work. > > Netscape is a known rogue client. Among other things, it does not > > cache the correct seperator character on a per "#xxx" namespace > > escape. So if you were reading news in "#news" (seperator "."), > > and went to create a new mail folder, it wouldn't use "/" as the > > component seperator when communicating the new forlder path to the > > server. > > Works with the Cyrus server, which is all I care about. > I don't think the imap spec hides enougth server implementation details, > which makes clients overly complex. Well, the Cyrus server doesn't handle namespaces other than the mailbox namespace, so it's not an issue. IMAP4 *does* hide a number of things from the client, like the requirement that the client implement MIME, or that the client implement a ".newsrc" for article "seen/not seen" marking in the #news namespace, etc.. The most annoying thing is the parenthetical presentation of headers in the "FETCH" command. Mostly, I'd say it hides different things than previous specifiactions. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199707011628.JAA18500>