Re: Using TM to help define the future of XML

Hi Patrick,

DISCLAIMER: I blame any and all typos on the iPhone and the train! ;)

On 2 Dec 2010, at 02:28 p.m., Patrick Durusau <patrick-Q/[email protected]>  
wrote:

> Andrew,
>
> Well, one top post deserves another! ;-)
>
> BTW, I would not get markup "history" from the xml-dev list. It is a
> mixture of myth, mis-information and mis-understanding. MMM instead of
> WWW if you will.
>
> XML was always about data as well as documents. You can search for Jon
> Bosak's comments on that point.
>
> The "data" argument is one of those, "now we are awake" and different
> from what went before sort of positions. Not based in reality or at
> least no shared reality of those aware of what went before. ;-)

The sentiments that led to the above were a mixture of many observed  
discussions on xml-dev, my own learning XML journey and my personal  
bias to more document-centric hypermedia formats.  However, I take  
your point. :)

This is also one area I see many less sophisticated developers  
struggle with as well. Extremely simplistic XML data representions are  
more the norm than the exception.

> [I deleted a long digression on markup theory and history. Of interest
> to few others and distracting from the point at hand.]

I'm always interested in history lest we repeat it, badly... ;)

> Markup (SGML/XML) made explicit certain information and encoded other
> information implicitly (unless you have a DTD/Schema). HyTime was a
> recognition that element names were simply markers for particular
> structures that could be transitioned to other structures. (ok, I am
> eliding a good bit, I want the post to fit into memory)
>
> So, yes, topic maps can expose the sort of information you are talking
> about, but only at the expense of its identification.
>
> I am not sure the "let's program to data formats" crowd is going to  
> see
> any advantage in the explicit identification of data upstream from  
> their
> use of it.

Representation formats are only relevant so long as they are useful. I  
think the heterogenous development envionments of today are driving  
more people to this idea as it is almost always required to serialize  
(convert or transform if you will) representations from one format to  
another. Much time has been spent writing the plumbing to make this as  
automagic as possible, but it still happens.

The problem is that the identity issue is generally ignored due to the  
safety of the traditional application/system boundary and the implicit  
identity space it defines.  I think the world has changed, and I don't  
think this can be an unquestioned assumtion any longer.

Rather than losing the identification of the subject described by a  
representation, I feel exactly the opposite would be possible.  Not  
only would you be able to link representations to the subjects they  
represent, but you can also explicitly identy both the semantics and  
its encoding rules using the same mechanism.

Anyone who has ever built or used data mapping tools should be able to  
enthusastically appreciate this expressiveness!

Why do you feel otherwise?

> Still, if there were a very robust and easy to use method for explicit
> identification and delivery of data into a variety of formats, it  
> could
> well be successful.

Indeed! :)

>>
>> --
>> Andrew S. Townley <[email protected]>
>> http://atownley.org
>>
>
>