[Nml-wg] Chameleon namespaces (was: XML syntax for NML relations)

[Jason Zurawski]

Hi Freek;

On 8/16/11 10:48 AM, thus spake Freek Dijkstra:
> Jason wrote:
>
>>> Even though its a special namespace, it can be reduced to the base
>>> which should make services happy.
>
> Freek wondered:
>
>> What do you mean with "special" namespace?
>
> Wait, you mean a Chameleon namespace? (you referred to
> http://books.xmlschemata.org/relaxng/relax-CHP-11-SECT-5.html earlier)
>
> I explicitly to NOT make that assumption. In fact, I'm very hesitant to
> use the concept of Chameleon namespaces, given the warning at the end of
> this page:
>
>> Chameleon schemas are very controversial. [...] Chameleon schemas
>> work contrary to most developers' namespace expectations, and in the
>> process remove most of the value of using namespaces. For this reason
>> I would recommend you be very cautious when using them!
>
> And this comes from the prime source on the topic...
>
> I do like the concept of subclassing, and I think we can (and should)
> include that (chameleon namespaces are a special case of subclassing).
> However, without proper consulting with a true XML expert, who thinks
> Chameleon namespaces is really the best solution to our problem, I am
> very hesitant to include that concept in NML.

The only advice I can give you is based on the experience of 
NMC/perfSONAR.  This is the basic design we have tried to follow over 
the prior years of design, implementation, and use.

It has allowed us to keep a basic structure specified, and yet allow 
extensibility to other data formats not envisioned in the original 
projects (e.g. new measurements can be specified, and implemented for a 
given service type very easily - we do not need to re-write the base 
each time this occurs).  To date the base has not changed in 5 years, I 
believe this fact alone points to extensible schemas being a good way to 
foster interoperability and encourage adoption.

Thanks;

-jason