Re: Summary: Green Turtle Status - RDFa Test Suite

On May 23, 2013, at 9:22 AM, Alex Milowski <alex@milowski.com> wrote:

> 
> 
> On Wed, May 22, 2013 at 11:23 PM, Gregg Kellogg <gregg@greggkellogg.net> wrote:
> On May 22, 2013, at 10:40 PM, Alex Milowski <alex@milowski.com> wrote:
> 
>> 
>> I'd love to submit an EARL with a 100% on all tests for all variants of XML, XHTML1, HTML4, XHTML5, and HTML5 but these 3 remaining tests cause me problems as I've detailed in previous e-mails.  I believe these are issues surrounding the test cases.
> 
> Sorry Alex, I ment to get back to you today on these issues; I'll look into them more tomorrow.
> 
> Note that we're really just looking for HTML5 reports, but it's good to have them all.
> 
> I'm trying to use the test suite directly without all the trickery I went through last time.  So far, I have succeeded in that I'm using a simple Jena-based service I build to test the outcomes.  It has worked well and I found some tests that weren't necessarily included in the JSON-LD manifest I was using previously.  As a result, I'll have an EARL for all the various test categories (i.e. XML, XHTML1, HTML4, XHTML5, and HTML5).
>  
> 
>> In summary:
>> 
>> Test 0109 / XHTML1 
>> 
>> This cannot be enforced in XHTML1 as xml:base processing will have already happened during most, if not all, modern XML processing.  Test 0109 isn't listed in the tests for XHTML5 and I don't believe it should be required for XHTML1.
> 
> For some reason, xml:base is _not_ used in XHTML1, but we did add it to XHTML5. If you're saying that the XHTML1 DOM model prevents this from being ignored, that's an interesting data point. Perhaps someone else has more perspective on why it must be ignored in XHTML1, and if tha is normative, or just the interpretation of a test; I don't think we can change any tests other than for HTML5 related specs at this point.
> 
> The point is that there is no such thing as an XHTML1 versus XHTML5 DOM.  There is just an XML DOM used for XHTML and so XHTML1 and XHTML5 get treated exactly the same.

Hmm, the processing rules for XHTML+RDFa 1.1 and HTML+RDFa 1.1 differ. XHTML1 uses the XHTML+RDFa 1.1 rules and XHTML5 uses the HTML+RDFa 1.1 rules, which do have some extensions for the XHTML case (namely xml:base), but also include the other differences for HTML5 processing.

> It is perfectly valid for a working group to change the test suite over time and most certainly has happened for other working groups.  I know we've had to adjust the test suite for XProc after REC to correctly match the specification or errata.
> 
> I don't think this WG should be in the business of enforcing XHTML rules for processing the xml:base attribute.
> 
> Again, my opinion, but this test is impossible to pass for a browser-based processor without doing things that would be considered "wrong" otherwise.  That is, I could try to "undo" xml:base but I really feel that would cause havoc for normal users who expected it to work because they put it into the document.

This is really useful input, as the intention is that a processor using the DOM should be able to pass the specs. From my perspective, I'd be happy to include xml:base processing in XHTML1. Similarly, if xml:lang cannot work for the HTML5 DOM making this non mandatory would be okay too. I don't think implementations should have to re-implement element language detection bacause of xml:lang. However, changing these will require some WG consensus.

> 
>> Test 0256 / HTML4 and HTML5
>> 
>> The xml:lang attribute is not recognized in HTML syntax documents and so the test will never succeed without violating the HTML5 specification.  The xml:lang attribute is not mentioned in HTML4 [1].  This test should be removed for anything in HTML syntax.
> 
> Note that there isn't an HTML5+RDFa spec, it's just been treated like HTML5+RDFa for most purposes, we could even just take it out of the test suite, but it is the only version of HTML that is a REC right now for which RDFa has any definition.
> 
> I wonder if something changes here, as I recall that @xml:lang as a non-namespaces attribute was to be treated like @lang; perhaps this is the peril of working against a living spec.
> 
> In either HTML4 or the non-REC HTML5, xml:lang isn't recognized in HTML syntax as the attribute we process.  I don't think anyone should expect xml:lang to work in HTML syntax as there aren't namespaces and so neither should RDFa.
> 
> IMHO, we should just remove this test as it just doesn't make sense. 

I agree, but we need WG concensus to change this.

> 
>> The manifests I'm using for these tests are:
>> 
>> XHTML1: http://rdfa.info/test-suite/rdfa1.1/xhtml1/manifest.ttl
>> HTML5: http://rdfa.info/test-suite/rdfa1.1/html5/manifest.ttl
>> HTML4: http://rdfa.info/test-suite/rdfa1.1/html4/manifest.ttl
> 
> What about XHTM5+RDFa? Do you intend to test that too?
> 
> 
> 
> All the tests for XHTML5 pass as the xml:lang attribute gets processed in XML syntax.
> 
> Here's my current status:
> 
> HTML5    195 / 196  - 0256 does not pass
> XHTML5  197 / 197 
> HTML4    167 / 169  - 0256 and 0303 do not pass
> XHTML1  180 / 181  - 0109 does not pass  
> XML         125 / 125
> 
> If I can sort these last tests, I can submit an EARL for all variants at 100%.  

It's the last 0.1% that takes 99.9% of the time :P

Gregg

> -- 
> --Alex Milowski
> "The excellence of grammar as a guide is proportional to the paucity of the
> inflexions, i.e. to the degree of analysis effected by the language
> considered."
> 
> Bertrand Russell in a footnote of Principles of Mathematics