Re: ID wording

[Tim Bray <Tim.Bray@xxxxxxx>]

On Aug 19, 2004, at 3:55 AM, Bill de hÓra wrote:

> Tim Bray wrote:
> > On Aug 18, 2004, at 11:49 PM, Mark Nottingham wrote:
> > > Ah. Can we agree to add this to the list;
> > >
> > > - If/when Atom is described with XML Schema, the ID should be a 
> > > xs:string, not xs:anyURI. I.e., the content will be restricted to a 
> > > URI in prose, not in the schema.
> > I'm nervous about getting that specific... this feels like it belongs 
> > in an Implementor's Guide or some such.
>
> Sure it's specific, but's a real point of breakage.

In my experience, RFCs which describe protocols rarely include 
discussion of the appropriate C or Java data types to hold things like 
IPv* addresses or DNS names. Now, you could argue that Atom is 
different from other protocols, or you could argue that schemas are 
different from programming language data types.  But it still feels a 
little weird to find this kind of implementation recommendation in the 
protocol description.  So...

> or, we just agree to remember that schema coming off this WG will 
> xsi:string.

I like that -Tim