Re: Identity modelling

Marshall Spight wrote:
> David Cressey wrote:
> > "Marshall Spight" <marshall.spight_at_gmail.com> wrote in message
> >
> > The fact that Roy wanted to know the context is interesting. If I were
> > asking the question, "is this URL in Google's master list"? Then it would
> > indeed be a key.
> >
> > My question was really about the relationship between the URL and the
> > resource. And I agree with you that this relationship is a pointer.
> >
> > Therefore the web should exhibit the drawbacks of designing based on a
> > pointer driven model. And it does, IMO.

Would you enumerate? I see it as a key-value approach or as a logical di-graph and it has issues such as "foreign keys" that refer to missing data (the key-value pair with that key is missing). When someone changes a URL, that does not change the sites pointing to it. However, this is very related to the cross-organizational, world-wide, nature of the web too. Can you think of a relational implementation of this highly distributed database that would be as successful?

As an aside, while there is a need for better tools to help improve quality, there is also something about this model (the di-graph) that makes for more tolerance in users. When I google and find even 7 out of 10 links with no data for me to view, I'm still satisfied with the three I get. Of course there is plenty of web-browsing frustration from users as well.

In addition to the traditional data quality concerns of accurate, timely, and useful data, there is also a component of quality that relates to acceptability, which in turn relates to process. Di-graph models like the web lend themselves to browsing. Driving gives a sense of being in control even if you encounter many dead-ends, one-way turns, etc. Being able to navigate through data by "turning" at a foreign key to navigate its target and from there to elsewhere in a web-like fashion yields a similar user satisfaction I think.

In systems based on the relational model, there might be (no solid data here) fewer processes built that give users this sense of being in control of how they "move around" in their data. Then again, I might be full of it on this one -- just thinking out loud. Did it resonate with anyone else?

--dawn

>
> Agreed on all counts.
>
>
> Marshall
Received on Thu Sep 01 2005 - 18:15:58 CEST