Re: RE : Question on Adding undeclared productions

On 11/14/2013 8:58 PM, FABLET Youenn wrote:
> Please find some comments inline.
> Regards,
>     Youenn
>
>> So, as an example, it is possible to have a non-terminal with productions like:
>>      ElemX : AT(*) ElemX    a
>>      ElemX : AT(*) ElemX    b.c
>>      ElemX : AT(*) [untyped value] ElemX   b.(c+1).d
>>
>> where the first two productions are completely interchangeable; whenever one could be used, the other could just as well >be used.  I don't see anything that prevents an implementation from actually using the longer event code (though I can't >imagine why they'd want to), so I suppose decode implementations are obliged to be ready to handle both the longer and >the shorter event codes.
> Encoders can do whatever they want but will generally probably choose the shorter version.
> Note though that one could tighten the rules so that only the shorter version is used .
> This is one of the rule expressed in http://www.w3.org/TR/exi-c14n/ (section 2.2).
>
>> I personally find this is a little surprising, so I think it would not be a bad idea to note the possibility of >interchangeable/redundant event codes in the text.  You might do so even in 8.5.4.4 as part of the introductory/overview >text.
> Note also that an implementation may well use a AT(*) production even though there is a more precise AT(qname) production available.
> The section that is supposed to deal with EXI event to EXI production is section 6, maybe it could be stated there from the beginning.

Thanks, Youenn.  I think a mention in section 6 is a good idea.

Regards,
Kevin

Received on Monday, 18 November 2013 15:27:42 UTC