- From: Markus Lanthaler <markus.lanthaler@gmx.net>
- Date: Thu, 6 Nov 2014 22:06:13 +0100
- To: <public-hydra@w3.org>
On 27 Okt 2014 at 20:44, Kjetil Kjernsmo wrote: > On Monday 27. October 2014 15.12.56 Markus Lanthaler wrote: >> What are the practical consequences of this? I think it boils down to the >> questions of what people will use hydra:totalItems for. Do you have an >> application that requires hydra:totalItems to be 100% accurate? > > Errr, no... :-) So, what it boils down to for me is that I don't see the use > for two vaguely defined terms. The approximate nature of void:triples is > presently useful. Clearly. And it will allow stuff like sampling algorithms > in the future. I just don't see the reason why totalItems should be roughly > the same, that just seems like duplication, URI aliasing, and a waste of > bandwidth to me. I much rather like to see it being defined as exact at the > time of the timestamp (which may be expressed as the Date header field in a > HTTP response). > > Just speculating: If you had a data stream management system (as opposed to > the database management systems that are usually underneath the stuff we > do)... Having the exact number of triples in a rolling or tumbling window, > which may be a small number, might be important and actually not difficult to > compute, no? > > So, I guess I'll turn the question around: Do you have any applications > that are able to make both terms useful if they are both defined to be > approximations? Not if we are talking about counting triples. Generally, totalItems however doesn't count triples (as in LDF) but items in a Hydra Collection (yeah you could also argue it expresses the number of ?collection hydra:member ?item triples for a fixed ?collection). -- Markus Lanthaler @markuslanthaler
Received on Thursday, 6 November 2014 21:06:38 UTC