Re: Moving towards "strict" xml:id checking

John Cowan a écrit :
> I'd rather see this:
> 
>    It is an error for an ID value to be specified more than once.
>    Processors MAY detect and report an error and MAY recover from it.

I feel uneasy about this.  This amounts to introducing yet another 
option in a technology with a stated goal of having as few options as 
possible (design goal #5 of XML 1.0).  And the underlying philosophy of 
Draconian Error Handling (cf. 
http://www.xml.com/axml/notes/Draconian.html) also comes to mind : one 
should be able to rely on the processor and not have to guess whether it 
found an error and ignored it, or failed to find it, or actually found 
that there was no error.

In short, we should decide whether ID uniqueness is important enough to 
mandate and stick to our guns, instead of trying to sit between two chairs.

As for which way to decide, that's another story.  My feeling is that 
the id in xml:id is an abbreviation of "identify" or "identity", and it 
seems strange to claim that a string identifies an element if any other 
element in the same document can bear the same "identity".

-- 
François

Received on Thursday, 2 September 2004 02:40:52 UTC