Re: machine readable properties

On Wednesday, March 19, 2003, 12:38:07 PM, Robin wrote:


RB> Chris Lilley wrote:
>> On Tuesday, March 18, 2003, 7:21:56 PM, Robin wrote:

>> RB> In the meantime I wrote a brute force Perl screen scraper (ah, that good old 
>> RB> '90s feel...) that finds regularities in the HTML of the spec to produce an XML 
>> RB> dump of the properties. I haven't yet checked that it makes much sense beyond 
>> RB> being WF. I've attached it in case anyone is interested (I can provide the 
>> RB> script too). Further parsing of the value spaces as well as recognition of 
>> RB> common options (yes, no, aural, N/A...) could be fairly easily added.
>>
>> Excellent, we should have a similar file available from SVG specs.

RB> SVG has some more complex things but is much more structured, it's certainly 
RB> doable. Why does that sound like an action item? :)

Please assign yourself a reasonable date to complete it.

>> The
>> only addition I could think of would be an xlink:href on the
>> percentages element, pointing back to said prose if the content is
>> 'see prose'.

RB> It's not possible to point to the prose directly because it can't
RB> guess where it is, but I can add an xlink:href to the latest
RB> anchor, which should be close enough (and in fact useful for other
RB> purposes).

Yes, I was thinking more for when the property list is used to
generate that part of the spec.

>> Next step is to maintain the properties in that form and generate the
>> (portion of the) html spec from that rather than the other way round.

RB> Given a spec published using XML and XSLT, a little trivial document() trickery 
RB> could easily replace <propdef ref='appearance'/> elements.




-- 
 Chris                            mailto:chris@w3.org

Received on Wednesday, 19 March 2003 11:36:18 UTC