- From: Sergey Melnik <melnik@db.stanford.edu>
- Date: Tue, 28 Nov 2000 13:58:52 -0800
- To: Seth Russell <seth@robustai.net>
- CC: RDF Interest Group <www-rdf-interest@w3.org>, Libby Miller <Libby.Miller@bristol.ac.uk>
Seth Russell wrote: > ... > In defense of Sergey's proposal, I would like to withdraw my allegation that > it would necessarily be less efficient for expressing context information. > As the examples below show, either method could be equally as efficient. > > My method: > [id7, context1, asserts, t1] > > Sergey's method: > [id8,t1, istin, context1] Seth, some clarification: as far as I can trace it, the first time the proposal to make statements intrinsic first-class objects by making them resources was (implicitly) made by Ron Daniel in the RADIX API by letting Statements subclass Resources. I inherited his idea in the "Stanford" API, have been relying on it for over a year for both in-memory and database-backed RDF applications, and it proved pretty decent so far. Thus, I believe it makes perfect sense to reflect this in the RDF model itself. Notice, however, that this proposal states nothing about how applications should use reification, e.g. whether they should use some notion of context, use statements as subjects or objects etc. etc. Whatever I used to back the idea served merely for illustration purposes, sorry if I caused misunderstanding. So, the two different representations above are nothing but developer's choice. In fact, our in-house database implementation uses what you call "My method" above, if you replace "asserts" by "contains"... Sergey
Received on Tuesday, 28 November 2000 16:40:48 UTC