>
> I think we would be talking about database identifiers rather than human
> readable database names - so this may not be an issue.
>
I had always imagined these would be database identifiers not names, so they
would be "well-formed" as far as communications and parsing is concerned. In
passing they are somewhat more than database IDs as they have to include
information (implicitly or explicitly) about the Host of the database and
other stuff, ([email protected] is not the same database as [email protected] or
[email protected] for example) but that is a detail, we call them Sources to
distinguish. The more 'interesting' aspect is that there is no standard for
constructing these IDs and certainly no registry of them, so a client
wishing to send a multi-database-search to two MSEs would have to
re-formulate the database IDs for each of them. This issue is likely to be
discussed at the forthcoming NISO IDs workshop.

Peter