[
Lists Home |
Date Index |
Thread Index
]
On Fri, 4 Apr 2003, Rick Jelliffe wrote:
> From: "Murali Mani" <mani@CS.UCLA.EDU>
>
> My point is that Henry showed something different: it applies only to DTD
> containing fixed or defaulted IDs or IDREFs. So while it applies to DTDs
> in general, it does not apply to *most* DTDs, and it is trivial to figure out
> if the DTD has fixed or defaulted IDs or IDREFs.
I concur. Henry's example was not proper use, the only thing is no one has
defined proper use so far.. I like ID/IDREF -- it allows traversing
relationships using path expressions.. Actually, we can convert any ER
model to XML w/o using any foreign keys, purely using ID/IDREF, which will
imply we can perform most queries w/o using joins, and hence lesser
mistakes in writing queries..??
Can I request you to take a look at
http://www.cs.ucla.edu/~mani/er03/modeling.pdf -- this is a quite easy to
read paper and should motivate very well, it is still under preparation,
though the ideas, trade-offs etc, I believe, have come out quite clearly..
> I have seen people use Henry's analysis as some kind of proof that ID/IDREF
> itself is bad, and I just cannot see the connection, if indeed it says nothing
> about normal use of ID/IDREF. (Who on earth uses fixed or defaulted IDs:
> it is very strange?)
>
> I think both the DSDL group at ISO and the W3C XML Schema group
> would be very interested in insight on how to move forward with keys.
I am focussing on this, and trying to understand what should be good ways
of doing things. I have sufficient encouragement and motivation to do a
good job. I hope something good will come out of our work..
> > More work needs to be done with respect to constraint specification.
>
> Yes indeed. Which is one reason I suspect the Schematron approach,
> of just levering what is convenient and comprehensible, is appropriate
> at the moment: if the properties of paths to trees are still active research.
cheers and regards - murali.
|
