Re: Serializations and data structures (was Re: Topic MapsonSQL)

len bullard (cbullard@hiwaay.net)
Tue, 15 Dec 1998 17:16:40 -0600

David Brownell wrote:
>
> Lars Marius Garshol wrote:
> >
> > As the argument stands it looks like checkmate to me.
>
> Because in fact the syntax and the data model are isomorphic
> in any good design.
>
> The disagreements come (IMHO) when the system isn't quite so neat
> as we'd like to see, so that the data model or the syntax have
> "extra" information. Which is the truth? This is a pragmatic
> question in my book. I tend to believe that specs get out of
> sync with reality in any successful system, so I've got a bias
> towards believing the actual data, rather than a version of its
> spec which is probably out of date ... :-)

I note that when creating an application of any kind, it is not
the design, the schema, or the spec that make bugs *emerge*:
it is the real world data. Witness the way a DTD grows like
topsy. The system is never that neat because the requirements
aren't. Good thing too. OTW, everything would have been built in the
last
century and we would all be steam engine repairmen.

I contend that topic maps, data types, and other ideas currently
circulating happen because we need them. Sort of a Duh, but again,
it points to the need to get requirements prior to design. I spent
the day creating a DTD for a language in which the original grammar
has very explicitly defined datatypes. I stuck CDATA everywhere
and am not very about that.

Of the various datatype proposals out there:

1. Which is most legitimate to apply today?
2. Which is likely to survive into XML1.>0?

I can use a stopgap if it has credibility. OTW, I put
these in comments.

len