Re: Suggested revised text for HTML/XML report intro

On Tue, 16 Aug 2011 16:09:00 +0200, John Cowan <>  
> The problem is that there is no compelling reason to prefer one approach
> to any other.

Of course there is. Processing XML requires no schema.  Processing XML in  
a lenient manner should not suddenly require a schema.

> Without such a justification, all we end up doing is
> complicating the description of XML further: instead of being able to say
> "report a fatal error", we must specify in detail exactly what infoset to
> produce for violations of each of the 83 productions, 12 well-formedness
> constraints, and 8 miscellaneous fatal-error specifications in XML 1.0
> (Fifth Edition).

In terms of complexity continuing processing or halting because of an  
error does not matter much. Because you have to check less character  
ranges a processor that just continues in face of errors might actually be  
less complex.

The idea that a tokenizer that does not halt in face of errors and  
produces output per a given data model is more complex is false. (HTML,  
WebVTT, CSS, event streams, XML5, are all testament to this.)

Anne van Kesteren

Received on Tuesday, 16 August 2011 15:55:31 UTC