14 May 2001 ERT WG Minutes

Summary of action items and resolutions



CR Concerned that need one techniques doc. If split between AU and WCAG to get techniques.

CMN There are techniques which are specific to, "how do I do this in a tool?" This is AU TECH stuff. AU TECH will incorporate tool as well as

Action CR: Talk with Matt May about HTML Techniques to make sure all AERT stuff being adopted there.


DD Which CSS test suite is most current, so that can generate EARL.

WC Who's that?

DD There are several. Talking with Chris and Bert. There are 4 or 5. All by individuals.

CMN If use the bookmarklet approach of Sean, we should be able to build one framework and no matter which test, should fit in. Interesting to say "here are john's tests and sean's tests, how many have I covered?"

DD Each test needs a URI that can be referenced. Annotation is good for tracking, but need to identify what gone through.

CMN If do it by page, get worse granularity, but something. One way to start is to take the bookmarklet and hook it up to Annotea so can collect results. Then go through the tests and say "yes or no." Need a way for bookmarklet to copy test in. Go to a test page, with a test suite, the bookmarklet could be written so that click "pass or fail". then ask your e-mail or other info or automate. add bookmarklet to browser run browser through thing. generate a bunch of bookmarklet. What attach annotation to? How to send it to annotea and how to get it back afterwards.

WC Need a field that will allow you to query for all annotations related to this test suite.

CMN Query by browser string.

Action SP: Work with Annotea folks to try to implement EARL bookmarklet.

WL Is annotea persistent?

CMN It's test data so doesn't matter if it dies. For our use case we don't need persistence.

WL Will it ever?

SP There are lots of suites that could output EARL. I was talking with P3P people, and I might hack an example for that. Did one for UAAG - a prose conformance statement that I expressed in EARL and sent to the group.

EARL 0.9

SP To implement need to have solid language. Need some updates, not to many. Want to use a lot of DAML in next version. Not sure about stability of DAML. Namespaces they use are datestamped.

CMN We and they are both doing draft vocabularies, I think we can run on whatever we get and see. If they stabilize, we can use, if not we'll write up the same properties.

SP Very primitive stuff in DAML that will always be in there - intersections, equivalents, etc. Not sure about date types. Should we have a standard date format? Couldn't communicate as easy if don't have one. Hard to constrain strings in RDF.

DD Is 0.9 usable? I'm reluctant to release 0.95 < 1 month. I rather wait for a couple months until we have 1.0 or implementation experience.

CMN It doesn't matter.

DD We announced to the world. We can't say we're playing with it.

SP There are some good improvements. I don't see problem making changes to 0.9.

CMN Should not change semantics in 0.9. No problem in having second namespace out there. Need to specify equivalences. Perhaps equivalents between 0.9 and 0.95 and say 0.95 better.

WC Better to do it sooner than later, before too many implementations. Also, 0.9 rather than 1.0 makes it sound more like a draft.

WL So have .9oops

SP The whole point of DAML/Semantic Web is to evolve languages and show how to translate from one to the other. There are still so many issues on the RDF issues list.

WC Anyone disagree move to 0.95. DD?

DD Ok. Keep old schema or replace?

SP All the old examples have 0.9, make new with 0.95. New namespace. Mapping should be in 0.9 or 0.95 schema?

WC Standard practice?

SP SWAG is to have elsewhere. If have 2 languages developed independently, won't know about each other. Logic is that all equivalent declared by 3rd party. But, might make sense to put in schema if you can.

WC What if someone finds 0.9, how do we encourage them to find 0.95?

SP Put a comment in 0.9 schema, or use "obsolete" so both machine readable and prose.

WC Any resolution for us?

SP This far out there. The problem is, there are more RDF people than semantic web people. Therefore, not a lot of best practices.

DD If we move to 0.95 we need to forget about 0.9. Perhaps people using w/out telling us. We should make clear that if people using, they should tell us.

WL A "current version" like documents?

DD But, it's machine-readable. Obsolete + comment. Also on home page.

Action SP: update schema

Action WC: update home page.

DD I have to leave. Still need a better tutorial. Entry stuff is not that easy to swallow. More readable than RDF, but it assumes a lot of background in namespace, context, subclassing, etc. Comments in example not enough. Should we have a tutorial? Introduce schema w/explanation.

SP I started writing one. It's so basic, to introduce each point, had to define 5 other basic points.

CMN Fair point to say how this works.

WL Earl for dummies.

CMN Right. We want pictures. Cut and paste examples. Pointers from EARL for dummies to EARL for not so much dummies anymore.

Action SP: send EARL basic tutorial to list

Action everyone: contribute to basic tutorial.

SP Every time I want to post something, I have to find a team member to ACL.

WC Could give you access to ERT space, there is date space there.

CMN It is good to collect stuff somewhere, we have a religious commitment to date space. ACL system basically sucks. If stuff you desperately need to get in and link to, either get scratchpad space in ERT or be nicer to push ACL guys to finish it, or enough so you can do own CHACL.

Action CMN: chase up ACL thing so that SP can ACL on his own.

WC Why are dates hard to use in RDF?

SP Can't have a standard date format. 2001-04-15 if next says 15 april 2001. That string used for date, should it be common between processes - "you must use YYYYMMDD" or leave it up to the processors and let them figure it out?

CMN Can't we use XMLSchema?

SP Possible. Not sure about the namespace. There's lots of uproar. It doesn't have # on end. In RDF to use types, must use #. If the 2 namepaces have any difference, not same.

CMN Why #?

SP If dereference, the data types in one document. All frag ids in one doc. TBL argues that a doc w/out # is a generic doc. With # means any former document. In XML Schema, no #. Therefore, going by namespace rec, the two are not equivalent.

CMN You can't start a name with a #?

SP nope.

WL HTTP handling with # is problem.

CMN # not part of URI.

SP Resource has diff meaning in URI spec: w/called reference. RDF resource is something w or w/out.

CMN Hack: can we declare a namespace by indirection?

SP Use something that shows namespaces are equivalent?

WC A hack b/c should have been done better, but isn't that the whole point of equivalences?

CMN Stick as open issue, but then find out how to do it.

SP Who do we ask?

CMN XMLSchema.

WL That leaves a "meanwhile."

WC Propose to use the hack and chase up XMLSchema.

WL But isn't date conversation to avoid this?

CMN Dublin date just says "use date." dc has subclassing. believe we can say "we're using date subtype of ISO format."

WC Otherwise, we can just say, "YYYYMMDD-TTTT"

CMN Regardless, want to type by Schema at some point.

WC Proposal: look at dc date, is there a subtype that we can use? If not, define own. Eventually use XML Schema to define.

/* SP reads from dc */

WC Time?

SP No, generic date.

Action CMN: chase dates so know how to use them and get consistency.

/* Chris leaves */

WL I think this project of making EARL available to others is dominant one. SP and Schlomit just had a interesting experience. She is having a hard to fitting EARL into reality, due to N3. One thing to decide is: do we push for N3 to become stable or expand to become real english. It is a valid project to be connected with EARL rather than ... take some load off of Sean. We need to produce real stuff.

CMN Do we have a canonical decision is it in XML RDF or N3?

SP I want to make sure that the syntax is as independent as possible. People see RDF as stable and N3 as not, since RDF a Rec. But RDF has holes. Therefore, both unstable. Therefore, need to be syntax independent, but need to represent it somehow.

WL I propose we begin representing in English.

CMN Begin by working from the model. EARL should follow RDF model. Then in a sytnax it works or we have an issue with the syntax.

WL We need to be able to explain it to me in English.

SP We need help from EO.

WL Perhaps time to go to them.

/* CMN leaves */

/* WL, SP, and WC go through EARL in n3 */

SP I can not get the keyword working in CVS. Just not doing it in N3. $id$

Action WC: find out why CVS keyword expansion not working in n3 files.

WL Right away, no clue.

SP Evaluation is the most basic thing in EARL. We have a list of them. Where is says "subclasS" that means that each eval has 3 bits. An assertor asserts an assertion.

WL Describes what an eval consists of: a statement which is a subclass of rdf:statement.

SP Statement says "this is a triple." w/earl eval, it puts specific properties at the ends of these. object assertor, predicate asserts, subject assertion.

WL Making your own vocabulary. The vocab of EARL is, one word is "asserts"

SP Yes, a core property.

WL Trying to find a way to put these in a TOC item. Vocabulary, property, ... what's best?

SP This is a data model, schema, ontology, vocabulary. In 0.9, this is not explicit, in 0.95 I found a way to change that. Greatest hack i've every done. Intersection of restriction of reification.

WC What is reification?

SP In EARL, say "sean says" then have another triple, "My page passes X" like being able to put a triple inside another triple.

WC Like containing?

/* any piece of triple can be reified, or all three. they can continue to reify infinitely. therefore subject, predicate, and object can all be reified, and reified themselves. */

WL Like language. Something so simple in concept can look so incredible.

SP I've got no comments. No real way that you can create your own model in RDF. In EARL we want to say a person asserts an assertion. I had to go to DAML. Had to restrict each of the pieces of triple.

WC What has DAML done that we can'?

SP Defined terms that RDF does not.

Action SP: Take question to DanC, RDF-IG or RDF-Logic. If not getting help, ask WC to chase up with W3C team.

SP I've asked Aaron, Libby, Seth. Aaron didn't know. Libby hasn't replied. Steph replied to one msg not the other. Could be a simpler way of doing it that I'm missing.

/* discussion about who doing semantic web stuff in Seattle. Discussion about SWAG. */

Action SP: send URI to list re: SWAG archives.

$Date: 2001/05/14 15:43:18 $ Wendy Chisholm