CellServDB contains cell entries (lines starting with '>') and host entries. The problem is that the entries in CellServDB are sometimes used to describe what cellnames exist, for example if we build a dynroot volume from that information. Other times the entries are only used to find the host lines for that cell.
The host lines in CellServDB which hold the hostnames and IP-addrs can be replaced by DNS lookups for AFSDB records. The big "religious" question is what should happen if the entries in CellServDB and DNS differ. Nsswitch.conf and other methods have been invented for host names. I think AFS should use the same prioritry order as for host names on the same computer. One special case of the problem above is a CellServDB with a cell entry for a cell but without host entries. On one hand we can interpret such an entry that the admin of that computer wants to indicate that there is a cell but the users should not access it as there are no DB server entries. On the other hand, there is the in my opinion more useful interpretation that there is such a cell, but the host entries should be looked up in DNS. Either that or invent a new way to register which cells should be shown in dynroot at startup. To use AFSDB and dynroot today, my cache manager needs a CellServDB with cell names only. With the implementation of the DB-server lookup in OpenAFS today, that will make vos and other commands that do not use the cache manager unusable for these cells. I think the questions was if that could be fixed. Harald. _______________________________________________ OpenAFS-devel mailing list [EMAIL PROTECTED] https://lists.openafs.org/mailman/listinfo/openafs-devel
