2008/6/16 Sam Johnston <>:
> The key thing with extensions is that if the spec allows for a feature then
> the output should be conformant (for that feature) rather than expressed as
> an extension, unless there is good reason to stray.
>
> A simple, quantitative metric that could potentially be applied is the ratio
> of in-spec to out-of-spec content (where implementations are punished for
> [ab]using extensions), but then there are questions as to whether you cound
> #directives, #bytes, etc.
Isn't that a judgement rather than a test Sam?
x:y is good, anything less is bad?
What's the pass fail test?
1.5 is a mess, basically untestable.
A review by the nvdl mailing list pulled quite a few weaknesses out,
which I've forwarded to the mainTC.
1.5 is also double crossed, referring to an OR clause for read/write.
Foreign is undefined.
I've been unable to define a clear test.
http://www.dpawson.co.uk/iic/tests.section1.html#test.section1.1.5.2.2
regards
--
Dave Pawson
XSLT XSL-FO FAQ.
http://www.dpawson.co.uk
