At 19:28 03-06-2003 +0200, Adam Dickmeiss wrote:

>PS: in order for this to work efficiently, the server must support HTTP

Cute. Of course it's *nearly* as good as Z39.50 concurrent operations, with
the major drawback (if you're dealing with potentially time-consuming
operations) that the order of responses has to match the order of the
requests... so if an operation towards the beginning of the list takes a
long time to execute, the user may experience a longer wait than he would
if the same operations had been executed in parallel, on separate connections.

Still, in the interest of keeping on friendly terms with a large
information provider, you'd probably see conscientious metasearchers
supporting this.

>Without pipelining, the server will process request completely
>and send response before proceeding to the next.

Actually, since the order is explicitly retained, many servers will
probably still do just that, since it's a lot easier than spawning separate
threads of operation to field one stream of requests.

Sebastian Hammer, Index Data <>
Ph: +45 3341 0100, Fax: +45 3341 0101