« Off to ISWC 2005 | Main | Up and Running with Nokia Research Center Cambridge »

2006-01-26

Is Wilbur Dead?

Some people have been asking about Wilbur, wondering if the project is dead. It is not. Lately, I have been working on persistent triple stores. The first prototype, up and running although just barely, uses MySQL with a "schema-ignorant" database schema. I am not a great fan of the idea of schema-aware storage models, although some folks seem to suggest that it would be a more efficient way to go. Perhaps, at least if you know the schema up front. I guess I am somewhat hesitant to start dealing with all the housekeeping that needs to happen when one makes changes to schemata.

After the MySQL triple store I was thinking of building a special version of the main-memory triple store that would run on top of AllegroCache. We'll see...

The reason why progress on Wilbur has been so slow is because I have been preoccupied with another project: Nokia has just started a joint research lab with the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL), dubbed "Nokia Research Center Cambridge". One of our projects focuses on the Semantic Web and the use of Semantic Web technologies to implement context- and policy-awareness. I'll report more once we make some concrete progress.

Posted by ora at 20:54

Comments

Have you thought of Elephant for persistence as a free alternative to AllegroCache? http://common-lisp.net/project/elephant/

I would also like Wilbur to be able to write files in RDF format out of its store, may be using some query filters

I've been using Wilbur 1.0+CMUCL for taxonomy manipulation in www.superpages.com and would like to thank you for the great system.

Posted by: Alec Segal at January 28, 2006 01:03 PM

1) Elephant is a suggestion worth considering. Thanks.

2) Serialization of triple stores is on my list of things to do, but I cannot say when I will get around to doing it.

Posted by: Ora Lassila at January 30, 2006 10:43 AM

Are there any known issues with the wilbur2 prerelease and sbcl?

I get many errors trying to load wilbur2 in sbcl on linux, the first of which is that "There is no class named KEYWORD", followed by many more.

Posted by: anon at March 20, 2006 12:15 AM

For some reason, SBCL does not recognize KEYWORD as a (sub)type of SYMBOL. This is a known issue which will get rectified in the next release.

SBCL issues many many warnings when compiling Wilbur. Most of these have to do with constant redefinitions that are harmless.

Posted by: Ora Lassila at March 20, 2006 12:20 AM

The issue with constant (re)definition in SBCL can be solved/fixed by wrapping the call to asdf in a hanlder-bind:

(handler-bind ((DEFCONSTANT-UNEQL #'(lambda (c) (invoke-restart 'CONTINUE)))) (asdf:oos 'asdf:compile-op :wilbur :force t))

My quick fix for the problematic KEYWORD class specifier in defmethod lambda lists:

(defmethod get-some-values ((frame node) #-sbcl (path keyword) #+sbcl (path symbol) (db db) index)

Not that this is pretty ... :-/

Posted by: R. Mattes at July 25, 2006 10:10 AM