Re: Version vs. Namespace

[Tim Bray <Tim.Bray@xxxxxxx>]

On Jul 12, 2004, at 9:17 AM, David Orchard wrote:

> While Mark's proposal of a profile might appear to be complicated, it 
> ought to get a bit more air time and explanation than the very next 
> message saying "no", especially with discussion of concrete use 
> cases/scenarios.

OK.  I claim that a single fixed namespace plus a single version 
attribute on the root element, plus an extensibility framework 
involving the usual mustUnderstand/mustIgnore primitives, will provide 
a nice efficient framework for the long term.  I further claim that 
there is now enough experience with this kind of versioning 
architecture that we understand the implementation issues.  I further 
express doubt that any further steps down the road to more complexity 
will have positive ROI.

So...... those who want a more complex and abstract versioning set-up 
should start, I think, from use-cases, so that we get a nice clean 
clear sense of exactly what the complexity buys us.  -Tim