Re: [xml-dev] RE: James Clark: XML versus the Web

[rjelliffe <rjelliffe@allette.com.au>]

 On Wed, 1 Dec 2010 15:51:17 -0800, Dimitre Novatchev 
 <dnovatchev@gmail.com> wrote:
>> XSLT is an event-based language so it ought to be possible to find a 
>> way to
>> make it handle user-input events (or data arriving from the server) 
>> by
>> firing appropriate template rules. But I've no idea how this would 
>> look in
>> detail.
>
> We just need a standard F & O function for this.


 OmniMark (the granddaddy event-based markup processing language of the 
 90s [actually
 a bit earlier ]) moved  to being more generic event-based IIRC. I don't 
 know how it
 worked out for them.


 Btw I think the OmniMark patent on "referents" will have lapsed by now. 
 It should be
 considered for the XSLT3 streaming extensions, since it was a very 
 viable technique,
 like an automatically managed "diversion" (was that the troff term?)

 Basically, in Omnimark you could have a string "hello %c world" and the 
 %c
 acts like XSLTs  <apply-templates />, processing the contents of the 
 current
 element. But if you could also declare a referent which was a stream 
 that
 didn't need to be closed before being used and was lazily evaluated. 
 The serializer
 would divert output to temp files when it encountered an unclosed 
 referent,
 and stall output until the referent had been processed. This allowed a 
 single
 a pass to include information not yet found. I guess an xslt3 streaming 
 mode is the
 nearest thing to a referent: however no decisions could be made based 
 on the value
 of a referent: it is a placeholder not a function as such.

 Cheers
 Rick Jelliffe