See also: IRC log
<chaals> If you have a range pointer (and anyway), you need to know whether pointers like linecharpointers refer to a range of one character/byte/etc or to a point just befre/after them (which is conceptually pure and beautiful but seems the wrong choice)
[discussion and refinement of pointers proposal by Johannes]
<scribe> ACTION: Johannes to send proposal to the list [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action01]
<CarlosI> do we need a pointer that points to a resource as a hole not to a part of a resource?, something like a ResourcePointer with just a reference property
<scribe> ACTION: Jim to read f2f minutes, consider proposals, and update the draft document accordingly [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action02]
Known Issues: http://www.w3.org/WAI/ER/EARL10/issues
Current WD: http://www.w3.org/TR/2006/WD-EARL10-Schema-20060927/
<scribe> New internal draft: http://www.w3.org/WAI/ER/EARL10/WD-EARL10-Schema-20070222
<chaals> ACTION: Shadi to say that recommended practice is to use foaf:name and one of foaf:mbox or foaf:mbox_sha1sum more clearly, and that foaf:organization should have foaf:homepage and foaf:name [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action03]
<chaals> Using other RDF (e.g. Dublin Core) to more extensively describe the organisation is recommended.
<chaals> ACTION: Chaals to write an example organisation [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action04]
<chaals> RESOLUTION: Drop example 7 - it is as confusing as helpful and we prolly don't really need it anyway.
<chaals> ACTION: Shadi, make the result have one outcome property [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action05]
<chaals> RESOLUTION: Change earl:instance to earl:pointer
<chaals> Should the pointer be restricted
<chaals> CMN: Don't think so, but recommend using pointer:pointer and making the ponter:reference be the earl:testSubject
CI: what about several test subjects? for example, testing consistent navigation on 10 pages at the same time, how to record 10 test subjects in EARL?
<JohannesK> <http://www.w3.org/TR/rdf-primer/#collections>
SAZ: either earl:TestSubjects, earl:WebContents, or >=1 earl:subject properties per assertion
CI & JK: don't like the latter, prone for missuse
solution: <earl:subject rdf:parseType="collection">
<earl:WebContent rdf:about="...">...</WebContent>
<earl:WebContent rdf:about="...">...</WebContent>
scribe:
</earl:subject>
RESOLUTION: describing multiple test subjects must be explained in the EARL 1.0 Guide
<scribe> ACTION: CarlosV and JK to implement this in the guide [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action06]
CI: can still be misused, for example to
summarize several assertions into one
... can not be checked automatically, must be interpreted by the use case
JK: how about using the dc:hasPart of the TestSubject class (to point to several sub-parts of a test subject)
CI: still have some concerns, need to think about it
<scribe> ACTION: SAZ to update example 7 accordingly (show the usage of multiple test subjects) [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action07]
<scribe> ACTION: CI to think about usage of dc:hasPart in TestSubject [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action08]
[discussion on earl:info -do we need a more elaborate warnings mechanism?]
general impression: info depends on the test case descriptions
CMN: leave the range of the info open so that people can subclass it or use it differently
<chaals> restricting it to Literal is a Very Bad Ideaâ„¢
SAZ: concerned about possible misuse, could add an earl:warning property when the time comes
<scribe> ACTION: SAZ think about the usage of earl:info and bring to the mailing list as needed [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action09]
<chaals> SAZ: Think about test subjects that are an http protocol - e.g. if you are testing serve headers, how do we do that?
[discussion about proposal to have earl:Content and drop WebContent & FileContent -see picture]
SAZ: source of the content is open, context is optional
example 1: for Web content, the source is either the actual response from the server or the URI if don't want to record the full content
the context is the HTTP interaction with the server that was used to retrieve the source
example 2: for file content, the source is either the actual file source or it is the file name
the context is the filename if the source is provided, empty otherwise
<chaals> you can make one now
CMN: source property is always a copy of the actual content, leave empty if not recording the actual content
SAZ: i am testing http://www.example.org/mypage.html, don't want to record the page in my report, what do i say?
CMN: don't put a source, put the URI in the
context
... the source is the content, the context is information about how you got
it
SAZ: in analogy, the same applies to file
content
... should encourage the usage of HTTP vocab for the context, even if not
recording the source or the body stuff
axtion: SAZ drop date property from the content class (that's all context stuff)
<scribe> ACTION: SAZ drop date property from the content class (that's all context stuff) [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action10]
RESOLUTION: keep the schema normative and the terms informative (to avoid any discrepancies -in case of inconsistencies, the schema in the spec (appendix A) wins)
<scribe> ACTION: SAZ remove foaf:Person from earl:SingleAssertor (its a subclass of foaf:Agent) [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action11]
SAZ: what if someone wants to define their own assertor?
CMN: subclass foaf:Agent, which can be pretty much anything anyway
SAZ: both earl:Software and earl:Content are not subclasses of earl:TestSubject, still we use them as testsubjects
CMN: does the property earl:subject have earl:TestSubject as its range?
SAZ: yes
CMN: then anything at the end of that property will be parsed as a TestSubject
SAZ: it only has to adhere to any TestSubject restrictions (currently there are actually no restrictions)
<scribe> ACTION: SAZ add "suggested types" column and add earl:Content and earl:Software there (as opposed to having them in "allowable types") [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action12]
<DanC_lap> I don't have any slides, but I'm happy to skype in and walk thru some web pages and code
<chaals> you want to show us stuff?
<chaals> (what is the question we are trying to answer)
<DanC_lap> dunno what question. this is just an experience report
<chaals> OK, that woud be great.
<chaals> wanna skype me?
<chaals> Or work through IRC..
<chaals> we are muted...
<DanC_lap> http://www.w3.org/2001/sw/grddl-wg/td/test_results
<chaals> waiting for FF
<chaals> ah, they got there too :)
<DanC_lap> a test: http://www.w3.org/2001/sw/grddl-wg/td/testlist1#projectsSpreadsheet
<chaals> 2
<DanC_lap> application/rdf+xml attachment: raptor_earl.rdf
<DanC_lap> http://www.w3.org/2001/sw/grddl-wg/td/Makefile
<chaals> DanC: We have two implementation reports in EARL. The raptor 1 is online
<DanC_lap> http://www.w3.org/2001/sw/grddl-wg/td/testlist1#earl_out
<DanC_lap> $ python testft.py \
<DanC_lap> --tester 'dan.rdf#dwc' \
<DanC_lap> --project 'grddlft.rdf#grddl_py' \
<DanC_lap> --run your_grddl_impl \
<DanC_lap> testlist1.rdf >earl_out.rdf
<DanC_lap> All tests were passed!
<chaals> DanC: If you tell it about you and the project it gives you EARL...
<DanC_lap> http://www.w3.org/2001/sw/grddl-wg/td/earlsum.py
<DanC_lap> and out comes the table... http://www.w3.org/2001/sw/grddl-wg/td/test_results
<DanC_lap> does earl have vocabulary for saying "target AA consists of tests 1 thru 6"?
this is part of a test case description language
yes, cannot tell
<chaals> CMN: No, we rely on saying "this RDF (== some result for AA) is equivalent to that result (== some result for each of tests 1 through 6)"
<DanC_lap> yeah...
<DanC_lap> <earl:TestResult>
<DanC_lap> <earl:validity rdf:resource="http://www.w3.org/WAI/ER/EARL/nmg-strawman#pass"/>
<DanC_lap> </earl:TestResult>
danc: the outcomes of a test are specified here: http://www.w3.org/TR/2006/WD-EARL10-Schema-20060927/#validitylevel
<chaals> I had convinced myself that you could do that in OWL, a few years ago.
<DanC_lap> so perhaps on the TestResult, I could just not put a validity, but put an rdfs comment a la "I don't claim to pass nor fail; I claim the test is broken"
<DanC_lap> owl doesn't have part/whole as a built-in
<DanC_lap> to say {a} = {b,c,d} is to say that b=c=d=a
<chaals> right. Say {a} = {b+c+d}
<chaals> ACTION: CMN to follow this up and figure it out again. [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action13]
<DanC_lap> the cost of namespace change goes up over time
<DanC_lap> it's OK to have a "subject to change without notice" policy around your namespace while you're working on it
<DanC_lap> or "subject to changes agreed in the ER WG"
<chaals> yeah, but painful in practice.
http://lists.w3.org/Archives/Public/w3c-wai-er-ig/2004Sep/0001
<chaals> ACTION: Shadi to make sure DanC is the first to know about namespace changes. [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action14]
<DanC_lap> ok, unless there are any questions; shall I ring off?
danc, see examples of OWL usage
http://lists.w3.org/Archives/Public/w3c-wai-er-ig/2004Sep/0001
<DanC_lap> (I just convinced myself of a reasonable way to say "the test is broken")
<chaals> DanC: The short answer is "it works"
danc, what's wrong with using earl:cannotTell as a result then use dc:description to say the test is broken?
<DanC_lap> if you coin "cannot tell" or "test materials disputed", then I'll probably use them.
<chaals> we have cannotTell
danc: the outcomes of a test are specified here: http://www.w3.org/TR/2006/WD-EARL10-Schema-20060927/#validitylevel
<DanC_lap> ah! news to me.
<chaals> but we are unlikely to coin "testDisputed"
<chaals> IMHO
thanks for experience report and implementation!
<DanC_lap> re 2004Sep/0001, I should probably take an action, but I won't be around when the action is tracked. I would appreciate a reminder in a week or two.
<scribe> ACTION: SAZ send DanC the OWL stuff [recorded in http://www.w3.org/2007/02/23-er-minutes.html#action15]