Problems about the Spectrum Data Model from the view of a Web Service programmer

[Gerard]

> 
> Yet again I find myself in the position of agreeing with Alasdair: there
> is nothing wrong with mixed content. In fact,  it is used elsewhere in
> the IVOA already (either in Registry resource records or VOEvent - I
> can't remember which offhand but have definitely come across it) so if
> we don't want to use it, we should be consistent right across the IVOA -
> a radical idea I know!

I vote for consistency and in particular would like us to be able to base
these on some agreed-to "good practices". The fact that something is used
elsewhere in the IVOA is hardly a guarantee of being good practice.
I have noticed it is hard to convince people of the usefulnes of good
practices (whatever they are). It is more difficult than getting the
specifications themselves, though we've had some progress there in the
Registry schema some time ago, and also in VOTable (not officially
implemented yet).

In the mean time, I had a question about the proper way to deal with mixed
content in schema definitions at the end of my earlier email
http://www.ivoa.net/forum/dm/0609/1169.htm 
A way of dealing with mixed content in an acceptable fashion (such as a
proper definition of the type of the content that is mixed in with the
elements) would go some way of persuading at least me of how to use it in a
proper way.

Cheers
Gerard

> 
>     Cheers,
> 
>     Matthew
> 
> Alasdair Allan wrote:
> >
> > Gerard wrote:
> >> I think irrespective of this, we might in the IVOA attempt to not use
> >> mixed
> >> content in serialisations of data and metadata simply beacues of
> >> examples as
> >> Alasdair mentions.
> >
> > Actually I was arguing precisely the opposite point, I don't really
> > see anything wrong with mixed content.
> >
> >> We can do without and it will be much simpler to
> >> transform between different representations (code, databasesof a given
> >> underlying model if every data element is explcitily named.
> >
> > Every data element is explicitly named in both examples I gave you.
> >
> >> Mixed content will allow things that are hard to make sense of in an
> >> automated fashion, like
> >>
> >> <tag1>34<tag2>35</tag2>36<tag3>hello</tag3>76</tag1>
> >
> > Err, no that's easier to make sense of in an automated fashion than it
> > is to read, although re-arranging your example into a more human
> > parsable version,  it's perfectly easy to understand as well...
> >
> > <tag1>
> >    34 36 76
> >   <tag2>35</tag2>
> >   <tag3>hello</tag3>
> > </tag1>
> >
> > In other words,
> >
> >     tag1         = [ 34, 36, 76 ]
> >     tag1.tag2     = 35
> >     tag1.tag3     = hello
> >
> >> Such mixed content is fine fore (X)HTML documents, but imho should be
> >> avoided if possible for serialisations that are to be read by machines.
> >
> > Why? I don't understand why you'd find the above ambiguous?
> >
> > Al.
> >