Re: Intent of ER-XML

On 2/26/2012 4:49 PM, Anne van Kesteren wrote:
 > On Sun, 26 Feb 2012 19:52:31 +0100, Noah Mendelsohn <nrm@arcanedomain.com>
 > wrote:
 >> Either way, I strongly feel that we should focus on the mapping, and not
 >> the processor or its API.
 >
 > Since this has been suggested a number of times now, I should probably
 > state that I'm not interested in working on this. I do not really see how
 > this would work, nor what the value would be.

OK, let's see what others think. I'm certainly not proposing to make a fuss 
or stand in the way of anyone doing good work on a processor spec. I think 
having something like that is probably far better than what we have today, 
but fwiw my intuition is that the layering of the specifications would be 
better if we first documented the mapping from input to output, without 
describing in detail any particular piece of software that might implement 
such a mapping.

 > I quite strongly believe that if we want this to work it has to be a
 > processor along the lines David Carlisle outlines in
 > http://lists.w3.org/Archives/Public/public-xml-er/2012Feb/0111.html so that
 > it can replace an existing XML processor.

That seems to imply that we have a specification for an XML processor. I'm 
not aware of any such specification. The XML Recommendation says a lot 
about how a string of input characters is interpreted as defining elements 
and attributes, and a few things about what processors must do (e.g. stop 
on errors), but it's not nearly enough of a specification to ensure that 
two processors in conformance with it would even come close to being plug 
compatible. Among many other things, it doesn't specify the form of output 
at all as far as I know.

FWIW, I'm fairly sure I've heard TimBL on a number of occasions express his 
feelings that the things the XML Recommendation does say about processors 
would better be moved into a separate processor specification.

In any case, your hard work on all of this is definitely appreciated. I do 
want to make the case for a mapping spec, because I think it's a better 
layering, but as I see I won't stand in the way if the consensus of others 
is to do a processor spec (or if you, as the person doing the detailed work 
at the moment, want to go that way.) Thank you!

Noah

Received on Sunday, 26 February 2012 22:21:55 UTC