-----Original Message-----
>From: Alex Osborne <[log in to unmask]>
>Sent: Apr 8, 2009 4:40 AM
>To: Peter Schlumpf <[log in to unmask]>
>Cc: [log in to unmask]
>Subject: Re: [CODE4LIB] Something completely different

>If I'm understanding you correctly, what you have in mind is a
>triplestore.  A database for storing purely relationship triples and
>nothing else.  A triple is made up of the name of an object, the name of
>a type of relationship and a name of another object.  A few example
>abstract triples might be:

Close.  What I have in mind looks more like just pure objects and pointers.  Think in purely abstract terms at this level.  It's possible to have one to many relationships from one object to many other objects.  It would be possible to group sets of objects together by using a single "proxy" object in this database that refers to the others.  But I'm starting to digress.

>> It allows for an interesting extension too -- weighting those
>> associations.  Suppose we use it to create a search structure, and
>> each time we go from one object referencing another we increment a
>> counter for that link by one.
>I'm a little confused by this.  What would the use case be?  Are you
>talking about a popularity score for each relationship to weight
>searches with?

Yes, that is one thing I had in mind.  It could be used to analyse browsing behaviour showing what is poular.  Such weights could also decay with time as they get stale.  But this is more an example of an application of the idea rather than something that describes the idea itself.


Thanks for your input,