Re: shape.dtd, diagram.dtd

[James K.Lowden <jklowden speakeasy org>]

On Thu, 11 Jul 2002 08:29:16 +0100 (BST), "Andrew Ferrier"
<andrew junk new-destiny co uk> wrote:

> > I have in mind generating Dia files from external sources,
> > "sql2dia" if you will.  Properly done, that would rely on the
> > DTD to determine the output structure.
>
> I'm not so sure that's true, is it? It depends on how much
> knowledge the generation/conversion utility has about .dia
> files, surely? If it was specific (as in the case of sql2dia),
> would not DTD validation just be a 'safety net' to ensure that
> sql2dia had not screwed up and accidentally created invalid
> .dia files? Surely it would have to be creating what it
> believed were valid .dia without reference to the DTD, unless
> it is going to read the DTD, parse it, understand it, and
> create files in accordance with it, in which case isn't it
> doing all the work of a DTD validator and more itself?

Oh, that's what I meant, more or less.  Suppose you want to create a .dia
file from scratch; what you have is a some data and an idea, but no Dia. 
Surely one approach would be to read the DTD, and use it with a validator
to verify your output.  Whether or not your output library actually takes
advantage of the DTD is a separate question, at least to me.  

--jkl