W3C home > Mailing lists > Public > www-rdf-logic@w3.org > June 2001

Re: Cheap memory [was RE: rdf as a base for other languages]

From: Jonathan Borden <jborden@mediaone.net>
Date: Tue, 5 Jun 2001 07:52:54 -0400
Message-ID: <088f01c0edb6$0f08a730$0a2e249b@nemc.org>
To: "Peter Crowther" <peter.crowther@networkinference.com>
Cc: <www-rdf-logic@w3.org>
Peter Crowther wrote:

> > From: Jonathan Borden [mailto:jborden@mediaone.net]
> > And realize that as
> > 256 Mb memory sticks have become quite cheap we don't need to
> > be as stingy
> > with tuple arity as 2 years ago. Remember that the Semantic
> > Web has lofty
> > goals and we ought not send it out in the jungle with dull knives :-)
> I take the other approach here and assume that two-year-old PC apps are
> current PDA apps are next year's WAPps --- it just depends what hardware
> you're running on.  We don't need to be *as* stingy, but if the Semantic
> is going to be used by agents then those agents will have to run on
> relatively small devices.

Certainly (my thoughts on this collected at:
http://www.openhealth.org/RDF/triples.html). One of my major arguments
against the current RDF reification system is the resultant explosion of
triples, regardless of the semantic implications. Similarly, the decision to
use n-tuples vs triples has a conceptual basis -- one can represent n-tuples
as a collection of triples but at the cost of transforming one _concept_
(the n-tuple) into many, (the set of triples). In an implemention, I imagine
these triples require more space for representation and more cycles for
manipulation. I strongly suspect that a relational db implemention based on
n-tuples will very much outperform an implementation using a transformation
to triples. e.g., imagine interating over a container using integer indexes
vs. searching for the pattern _nnn, converting the string "nnn" to an
integer etc. When every rdf:Description has an associated bag, this is a
real issue.

This very much reminds me of the CISC vs. RISC wars of the 1990s. When
memory was expensive RISC lost out because of increased storage
requirements. So real world efficiency is a hard thing to predict.

Received on Tuesday, 5 June 2001 08:09:58 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 2 March 2016 11:10:35 UTC