On Fri, Sep 26, 2003 at 09:15:52AM -0800, Greg Newby wrote:
> I think that we'll need to think about scalability of GridIR,
> particularly with regards to network traffic and latency.
> But that is not something we need to work on immediately --
> better to get the basics working, so we can see how the
> traffic patterns look.
One related point which we touched on today was that Z39.50 is
stateful (or maybe more accurately, as Ray pointed out,
connection-oriented but not really very actively stateful), but SRW is
definitely not stateful. Now, since GridIR will be stateful anyway, I
was wondering if there is anything to be gained by taking advantage of
that fact (in the context of this discussion of SRW); and if so, would
SRW need to be transformed in some way to do that? I know that's
quite vague, but I wanted to put that on the table in case it might be
worth considering. In more general terms (that is, even more
vaguely), has SRW been simplified from Z39.50 in ways that we (GIR) do
not need to be simplified and which we might prefer not to be
simplified (for example, because SRW was specifically intended for
implementation in web services). That is a question that suggested
itself while I was working through some of the SRW documentation.