W3C home > Mailing lists > Public > semantic-web@w3.org > July 2007

Re: RDFON: a new RDF serialization

From: Anil Lotek <_@whats-your.name>
Date: Thu, 26 Jul 2007 20:21:48 -0400
To: Semantic Web <semantic-web@w3.org>
Message-ID: <20070727002128.GA13870@replic.net>

>  My understanding is the decision with SPARQL was to make it a) SQL-like,
>  and b) similar to N3 (as Tim just noted). And it has XML and JSON results 
>  formats.

sure.. SPARQL seems like its awfully powerful (from Andy's PDFs), especially to big-pharma biotech DNA hax0rs and big-brother dataminers. not so much the average web user or developer though..who already has plenty of nice ORM tools at their disposal, where they can write queries devoid of the extra 'thinking' step required to figure out what a sparql query does..

>  Seems reasonable enough to me.

if SPARQL is going to interest those people (real world devs of the web, who have the potential to bring on the 'semantic web' by sheer deployment numbers and influence), it has to be even simpler. and it isnt, because,

they'll need a JSON->SPARQL-syntax transformer. an additional layer to write, maintain, potentially fail.. this is no better than their existing ORM, just 'different'. itd be trivial to assign FOAF/SIOC/etc URIs to ActiveRecord object predicates, and voila, all Rails sites join the seantic web. yet nobody has claimed to trying this - intead we've got useless crap like ActiveRDF that adds an additional layer of slowness when the native Hash type already can store RDF just fine and has a more useful native toolbox of manipulation methods.. likewise ActiveRDF implemented and primarily supports its own SQL backend putting itself somewhere in the 'un-critical-mass failed ActiveRecord competitor' category.. the ill notion here being a 'new stack' (sorry, 'layer cake') is necessary to get semantic - the real-world stacks are continually evolving, i doubt they will intersect with n3 or sparql - or even new classes (uris work fine as property names in most sane languages, we dont need bloated memleaking stdlib-obcuring fiascos like Redland via SWIG)


also the JSON results format (documented on w3.org) is an impedance mismatch hack from hell. the returned data also isn't idiomatic to JSON (ie, 1 resource == 1 object). 

instead of just flaming, i'll ask why there still isnt a writeup on serialization where rdf resources == json objects. or the fact that CONSTRUCT with just a where clause (and implied construction of all resulting triples) into this format is a lot less of an impedance mismatch hack from hell. heck, doing a plain SELECT and reconstructing yourself using the binding name as the predicate, row as the subject and the tabular cell value as the object is less of a hack from hell.

Anil
Received on Friday, 27 July 2007 04:22:50 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:58 UTC