Mailing-List: contact jackrabbit-dev-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: jackrabbit-dev@incubator.apache.org
Received-SPF: neutral (asf.osuosl.org: local policy)
Message-ID: <42E968CB.1070709@apache.org>
Date: Thu, 28 Jul 2005 16:22:51 -0700
From: Stefano Mazzocchi <stefano@apache.org>
User-Agent: Mozilla Thunderbird 1.0.6 (Macintosh/20050716)
MIME-Version: 1.0
To: jackrabbit-dev@incubator.apache.org
Subject: Re: associations between nodes
References: <BAY101-F21B6F57813C83F4C931F6CBFCF0@phx.gbl>
In-Reply-To: <BAY101-F21B6F57813C83F4C931F6CBFCF0@phx.gbl>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

James Myers wrote:
>> hi jim,
>>
>> > Prof Mazzochi sounds 'way off' ;-) ,
>> well, so be it ;)
>> i just thought that something like this could address michaels idea
>> of the bi-directional references.
> 
> Absolutely - the example was fine, I was just concerned about Stephano 
> being allowed to 'officially' mess with students'minds :-)

I'll continue to do it unofficially then ;-)

>> > but thats how I've seen references - as the way to encode RDF. I 
>> forget how
>> > far we backed off in the final spec, but the idea of including a 
>> "following'
>> > clause in the query language was a way to let you query along graphs.
>> agreed. with respect to query languages we certainly have the liberty
>> to extend beyond the spec in jackrabbit. but maybe jcr:deref() is what
>> you were looking for.
> 
> 
> I think that's where consensus led us - jcr:deref allows you to 
> 'manually' traverse references to find things. The 'following' clause in 
> the query language was intended to let you substitute named reference 
> relationships in for the parent/child relationship in search, so one 
> could formulate a query along the lines of "find all works where creator 
> = bob that are 'derived from' /a/b/c" where 'derived from' would be a 
> reference property and you'd want to recursively dereference nodes 
> linked by 'derived by' looking for those created by Bob. While useful, 
> there were/are valid concerns about this being 'expensive', 'following' 
> being nonstandard, etc. so it was pulled out with the possibility that 
> it could be added back in an alternate query language.

If we find a way to encode 'named' references, it would be possible to 
write a Sparql implementation in JackRabbit and turn it into a triple store.

It if scaled and wasn't that hacky, it would make a lot of people happy, 
especially because there is no standard API for adding and querying 
triple stores in the java world (and there are at least 4 
competing-yet-similar APIs)

At the same time, I'm not holding my breath and I don't have time to 
make this happen.

-- 
Stefano.