The syntax of the CellServDB file format [1] is well-known but is it a
standard from the perspective of this group?  The answer to the question
will determine where discussion of replacing the file format should take
place.

The current format is limited in what can be described and is
unfortunately interpreted differently depending upon the client or
server that is in use.

For example:

 . servers only examine the IP addresses in the file
   whereas the clients sometimes examine the IP address and
   sometimes examine the DNS hostname

 . some clients treat a cell name followed by an empty
   server list as an indication that DNS lookups should be
   used and others do not

 . some clients understand server clones and some do not

 . some clients support linked cells and some do not

Here is some of the new information that I think should be included in a
future replacement with per cell granularity:

 . dns = {always, prefer-dns, prefer-local, never}

 . default port values for afs services

 . optional IPv6 addressing

 . local Kerberos realm list

 . service specific configuration data

 . server specific data

   - default rank

   - clone status

Should this discussion take place as a standard or should it be treated
as an implementation specific format?

Jeffrey Altman



[1] http://docs.openafs.org/Reference/5/CellServDB.html

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to