> In response to Matthew: it is not explicitly part of the
> SRU/SRW protocol that referencing a resultset can generate a
> new resultset with a new name.It would force clients to
> refresh the resutset after each access. I do not see a reason
> to do so; when there is a resultset why make a new one that
> is the same? I can imagine that the server uses each access
> to a resultset to increase the TTL.
Its not explicitly part of the protocol that it wo'n't either.
There was some discussion at one point (I think in Washington last year)
as to whether a server would return the same name with an updated TTL or
a new one. I thought the consensus was the latter, although it might
have been that both were allowable behaviours.
I don't see this as a major problem - we aren't generating a new result
set per se (so the client needn't refresh) just a new identifier for
that result set. If both identifiers are within their TTL both are valid
for use, although it would be advisable for the client to use the latest
identifier for referencing the result set. A server may wish not to
maintain multiple identifiers for the result set and hence may prefer to
return the same string with a new TTL.
However, if it is felt we need to specify one or other of these
behaviour to avoid confusion then fine.
Some of the comments have made me feel that ResultSetName is a misnomer
and we should talk of result set identifier?
Matthew
|