- From: Seaborne, Andy <andy.seaborne@hp.com>
- Date: Thu, 27 Oct 2005 10:10:28 +0100
- To: RDF Data Access Working Group <public-rdf-dawg@w3.org>
Jeen Broekstra wrote: > Kendall Clark wrote: > >>On 16:54, Wed 26 Oct 05, Jeen Broekstra wrote: >> >> >> >>>FWIW I do not think the current design to make *excessive* use of >>>bandwidth, except in corner cases. YMMV. >> >> >>Well, FWIW, that corner case is the commentor's *common* case; that >>is, the primary of SPARQL and the results format involves >>(necessarily) lots and lots and lots of unbounds, because the queries >> involve OPTIONALs and UNIONS (however they're spelled, I can't >>recall). > > > Yes, you are right, I should not have put it that way. I only meant that > for many use cases/queries (in which optionals and unions are not, or > sparingly) used, the bandwidth use is just fine with the current design, > and for many use cases, the ease/speed of processing is a major issue. > > >>That's precisely the problem that led to the unusual step of thinking >> we could save some bytes and still have an easily human readable >>format, keeping it in XML. >> >>I don't believe the change to <binding/> makes processing that much >>more difficult. > > > To clarify: I think my major concern (with both stripping and > collapsing, though mostly with collapsing) is not so much that it is > more difficult but that it is more costly in terms of processor > performance. Note that Ron's tests only give performance figures for the > case where you actually _have_ a significant size reduction (because it > contains a lot of unbounds). > > I'd like to see some figures on how comparative processor performance is > for result sets that contain no unbounds (I'll see if I can come up with > some figures on this myself tomorrow, or perhaps Ron will do this, > you/Bijan mentioned he'd do some extra tests). That would give us a more > complete picture of the consequences of either design. > > >>>If, for purposes of minimizing the result set size in bytes, we >>>offer a binary format with the reduction in size and processing >>>time mentioned above, I think that would address his concern, >>>although of course such a format is can not be processed with XSLT. >>> The other option of using GZIP compression is still a viable >>>alternative as well, IMHO. >> >>I am only guessing here, and Bijan mentioned that Ron will be doing >>some further tests, but I'd be really surprised if our organization >>got behind a binary format. But, again, that's just a guess, not a >>position. > > > I can very well imagine that it is simply not a good idea to introduce > this extra format into the WG at this stage of the game. I am happy to > invest time into it though if other people think it's worthwile. > > *shrug* we need to document it for Sesame users anyway ;) > > Jeen > A different perspective: ARQ/Joseki3 streams both client and server. With HTTP, processing of results in the client can proceed in parallel with the server generaing them. The time-to-first result is now independent of the transmission and XML parsing costs of the whole result set. The time-to-last-result is lessen because the the client processing costs are in the time dead, wait time that would occur in a serial pipeline. In addition, the client end can now use StAX, a pull parser, so it is also using a lighter weight parsing system than XSLT processing over a full DOM. This could then be connected to STX for streaming transformation of results. For SOAP, I haven't done this yet. SOAP stacks are being to appear with the possibility of streaming so I'm hopeful here as well. > *shrug* we need to document it for Sesame users anyway ;) You document it and I'll (look to) implementing it :-) My requirements are as for the discussion had pre-DAWG; I'd like to be able to process the results without having to deeply inspect the query so a declaration of results variables helps. And a MIME type. Andy
Received on Thursday, 27 October 2005 09:10:36 UTC