On Thu, 2016-01-21 at 14:02 +0100, Carlos Garnacho wrote:
Yeah, afair the problem with Apples Spotlight RPC protocol is that it has no way returning results in chunks and tell the client to come back immediately for more results. Apples Spotlight RPC server simply puts *all* *available* (this depends on how many results the first async query returned) results in one RPC packet (up to some limit imposed by SMB/RPC).Oh I see, yay design limitations :). Thanks for sharing.
People working at big companies making RPC APIs that are not cursor nor page based. They'll learn from the many catastrophic API mistakes ... ... slowly Better is to have these protocols and APIs to be developed in the open, for example at IETF. That way smart people would have reviewed and commented on it. Unfortunately do many big companies have a huge NIH syndrome. On top of cursor and/or paging based, it would need to do pipelining and allow live updates on top. So "remote procedure calls" is not a good way of doing such protocols in the first place. Ah well, slowly, very slowly, they'll learn. I guess some day we'll need to just make a protocol for metadata sharing between SPARQL endpoints, or maybe just do this: https://www.w3.org/TR/rdf-sparql-protocol (hmm, over soap, RPC-style and http. that's typical for w3 and ... bad). Kind regards, Philip
Attachment:
signature.asc
Description: This is a digitally signed message part