Re: Additional syntactic restrictions

[Gavin Thomas Nicol <gtn@xxxxxxxx>]

On Friday 07 June 2002 09:14 am, Chris Lilley wrote:
> GTN> I would argue that this is a faulty implementation if the payload in
> this GTN> hypothetical protocol is supposed to be arbitrary XML content.
>
> Yes! Exactly. It would be faulty. It would be a bug. it would not be
> fully interoperable. QED. Re-use of an apparently stock, but actually
> modified, XML parser would have affected interoperability.
>
> >> As i am sure you are aware, if you make a claim of impossibility it
> >> takes only a single counter-case to disprove your claim.
>
> GTN> Sure, so long as the claim isn't based on assumptions of incorrect
> GTN> implementation.
>
> If you say that something is impossible, and I show that it is
> possible, then I disproved your claim. It doesn't matter that I used
> an example of a bug in software. Software does, I am sure you are
> aware, ship with bugs. The possibility of this sort of bug being
> introduced is sufficient to dissprove your claim.

Yeah, whatever. You can split hairs if you choose to. I would contend that 
it's not the protocol, or the parser, that affects interoperability, just the 
guy implementing the spec that's too lazy/ignorant to conform to it. That 
kind of laziness, and catering to it, causes/caused more grief that anything 
I'm aware of.... in a case like the above, I'd say "Hang the engineer!".

> GTN> Yes. Do you have a specific case in mind where you'd have an XML
> protocol + GTN> an XML payload in mind that wouldn't somehow involve a
> restriction on the XML GTN> features allowed in the protocol-specific XML?
> Perhaps you are arguing that GTN> protocols should accept XML, period? My
> assumption is that the XML used for GTN> the protocol, and the XML used for
> data are logically separate entities and GTN> would be processed as such.

These are the more interesting questions that you conveniently didn't answer. 
If I have any protocol that uses XML, and in any way subsets it (including 
use of CDATA sections, external entity references, subsets, etc. etc.) they 
fail your interoperability test in precisely the same way you claim my logic 
above for UTF8 onloy did, because in order to implement the protocol, a 
subset parser is sufficient. This is *precisely* the same logic used for the 
Unicode document character set proposal made long ago, and which has since 
found it's way into both XML and HTML.

Is your argument that XML-based protocols should accept arbitrary XML, 
period?