ERT WG

Fuzzy Pointers

saz: jim sent descriptin http://jibbering.com/discussion/fuzzy-pointers.html, first concept of fuzzy pointers

saz: we'll probably not be able to note all document changes

<chaals> Chris: this is astrategy we can use for things that are not well-formed. sounds like it would be ok. will we use other things as well, such as xpointers?

cr: fuzzy pointer for docs that are not well formed ok

<Zakim> chaals, you wanted to say good thing overall

<shadi> Chaals: Fuzzy Pointers are compatible with XPointers

<shadi> Chaals: FP is one strategies, we probably need more than one

<chaals> chaals: we will want several Locations at the same time, to give us good chance to find things

<shadi> example: /html/body/table[2]/*/tr[1]/td[0]/img[@src='moomin.gif']

saz: example with tbody, is possible to express with xpointer

<chaals> SAZ: is it not possible to use an Xpointer for something in a table that is invalid becausee has no tbody (see example above)

<chaals> jim: It is possible, but less accurate, because you could have any number of things replace the * whereas with fuzzy pointer you know where the thing was.

<chaals> saz: we could use this to have a very rough pointer.. such as any images with teh following attribute...

cmn: fuzzy pointers are good thing to use, need examples

<niq> "this passcode is not valid" (many, many times). And when I ask for operator assistance, it's a recorded voice telling me to wait for an operator (but no end to the wait)

<chaals> chaals: propose that we include an example of how to use a fuzzy pointer as a location in next spec draft.

saz: what is the practical experience with fuzzy pointers

<chaals> jim: Was looking at it in the context of anotea

<chaals> ... got a lot of reliability across browsers for identifying things in browsers

<niq> fuzzy pointers: arises from when we wanted Jim's client code (various) to interoperate with my server stuff (Valet)

<chaals> ... You could define a similar pointer for cases like images - we should do some more work on how they would interoperate in earl context

Proposed Changes to the EARL Schema

<shadi> http://lists.w3.org/Archives/Public/public-wai-ert/2005Apr/att-0212/schema.html#prop-assertion

saz: addition of rdf-id?

<Zakim> chaals, you wanted to exaplin the relevant bit of http://lists.w3.org/Archives/Public/public-wai-ert/2005May/0012.html

<chaals> cmn: not sure that it should be required to have an ID, although it is a good practice

cmn: do we really require it?

<chaals> cr: what is the purpose of it?

<chaals> saz: if you have a few assertions, and ask for the assertion with specific property, you get the assertions merged together

<chaals> cr: can't you jsut ask for information that differentiates the assertion?

<chaals> saz: there are sometimes multiple assertions with same subject, same result, same testcase, different tools...

<chaals> cr: there should always be something different in the assertions. I don't see a case where you need an id to differentiate them. Or is this to allow you to refer to them?

<Zakim> chaals, you wanted to give use case based on proposed evidence property

cmn: can't refer to a specific assertion without id

<chaals> cr: how do you generate the ID?

cmn: id tied to e.g. the URI of a page, no mix-up

<chaals> jk: if you merge two things with the same ID, what do you get?

<niq> twins? :-)

<chaals> cmn: this should be handled by RDF parsers...

<chaals> jl: I don't think it is bad, but I don't think we should make a comment in the schema to require it - this is more a good practice

<Zakim> chaals, you wanted to say "what jim said"

<chaals> saz: I thnk we shouldn't reinvent wheels reproducing vocabulary

<JibberJim> Rosco - http://swordfish.rdfweb.org/discovery/2003/08/validation/

<chaals> ... but do we want to have a minimal set of required data in earl

<chaals> ... and include in that an ID

<chaals> cmn: we should have a set of required information for an EARL assertion

<chaals> ... not convinced that ID should be part of that

<chaals> cr: should have minimal set...

<chaals> saz: question is does ID belong in that set. let's leave for later.

saz: rdf-id good think, optional or required not clear by now

<chaals> ACTION: cmn to look up how to define ID as required [recorded in http://www.w3.org/2005/05/03-er-irc]

<chaals> jim: the link above is a checker that says whether a FOAF file has enough useful information to be interoperable. that can be a more helpful approach than formally defininf rules in the schema

<chaals> ... http://swordfish.rdfweb.org/discovery/2003/08/validation/

<chaals> ==

<chaals> location

saz: next property location

<chaals> saz: pointer to a container of Locations

<chaals> cmn: good idea, only suggestion is that the URI for class and property look different

<JibberJim> classes are pascal case - properties are camel case I believe

saz: next property evidence
... based on the properties listed

cmn: shouldn't be restricted to assertions, rule is not an assertion necessarily

saz: should the range change?

<chaals> jim: don't see that removing the restriction will lose a lot of interoperability

cmn: expect a list of assertions + rules why these are relevant
... most tools don't need to expect all the evidence
... without a rule we're providing random statements, should be possible to include reasoning

thx to all for helping out :