I can see that it would be a problem but the issue is that the NFS spec cannot control the underlying file systems that the NFS servers use. In most cases, this is not under the control of the server implementor. So, these days, it would be unlikely for a file system not to support double-s, but there are other less common cases where the same sort of thing can happen. A file system may not support characters beyond 0xffff and the spec defines that those should result in NFS4ERR_BADCHAR. That's the way v4 handles the issue and you can argue that we should be more demanding of file systems, but you are going to get a lot of resistance if you propose that for v4.1, for example. A similar issue is that file systems may make certain names invalid. Unix file systems generally don't let you create the name ".." since it is reserved for an upward pointer. If you use the name "..", you will get NFS4ERR_BADNAME. However, on some systems you may be allowed to create files with such a name. That is no guarantee that you will on all systems. Servers might abuse the discretion that spec allows them by rejecting all names with a prime number of characters (for example). They would be compliant (officially) but nobody would use them. Is that a big deal? -----Original Message----- From: Patrik Fältström [mailto:paf@cisco.com] Sent: Friday, June 20, 2003 11:53 PM To: Noveck, Dave Cc: Harald Tveit Alvestrand; spencer.shepler@sun.com; beame@bws.com; brent.callaghan@sun.com; mike@eisler.com; david.robinson@sun.com; robert.thurlow@sun.com; Mark Davis; IESG Subject: Re: RFC 3530 On lördag, jun 21, 2003, at 05:44 Europe/Stockholm, Noveck, Dave wrote: > The actual users are a different matter. I certainly see that > they could be quite legitimately discomfited if the mapping is one > that they do not expect. They might also find unusable a server > which simply rejected any names with a double-s character with > NFS4ERR_BADCHAR, which would be compliant with the spec, but > not very useful for a German speaker. The problem would be if a user have an NFS server which do handle double-s. Then they switch vendor to another NFS server which claim to be compliant with the spec. They move the files there (not using NFS) and suddenly some files are not accessible from the same client. That is non-interoperability to me. paf