OWL Working Group Meeting Minutes, 04 June 2008
DRAFT. Currently Under Review
See also: IRC log
- Bijan Parsia, Sandro Hawke, Evan Wallace, Peter Patel-Schneider, Boris Motik, Uli Sattler, Michael Smith, Elisa Kendall, Michael Schneider, Achille Fokoue, Markus Krötzsch, Ivan Herman, Diego Calvanese, Ratnesh Sahay, Alan Ruttenberg, Rinke Hoekstra, Carsten Lutz, Zhe Wu
- Alan Ruttenberg
- Elisa Kendall
(Scribe changed to Elisa Kendall)
(No activity for 37 minutes)
you put it into one of our reasoners, you see whether or not it can be done ... let's just do this
there is different motivation for these, for keys it's a huge wart on the language, but we can request implementation and see what happens
so there is always a wide balance of considerations, and no one was trying to suggest otherwise
whether we keep this on a separate page or include them is a different issue now -- we're at the implementation stage and need to see whether or not we can do them
Alan Ruttenberg: would there be any strong objectors to dropping top and bottom role? do you think there isn't any reason to spend more time on them?
If you say yes then you really want them
Alan Ruttenberg: so it's been mentioned several times, so regardless of whether we put something into the spec, we will need 2 implementations
Alan Ruttenberg: so I'm wondering whether we should resolve such problems by saying that we should put these things in and see what happens, or not
the advantage of putting them in is that folks have something to think about for longer
Sandro Hawke: the W3C key is to say that something is at risk, then if you take it out later you don't have to worry about the process
Boris Motik: I don't think that discussing this over email would be useful -- my proposal would be to implement these features and then come back and say yes this was the experience
my proposal would be to postpone this for a week, 2, 3 and then see what really happens ... in my case the implemetnation isnt really there so it would be a month before I could come back with an answer
Alan Ruttenberg: we should discuss next week whether or not we should put things that are at risk into the spec
Boris Motik: I also looked in one of these RFCs, regarding "should" and i wasnt happy with that because it says this is optional, and I would like something more than optional
we've been using this to say something is default ... do it like this unless you have a very good reason for not doing it
Bijan Parsia: it doesn't mean optional ...
Boris Motik: then I looked at a different document
Bijan Parsia: the official definition is what you were asking for, but in general, "should" can have the effect of being optional, or could have the effect of being mandatory, depending on how you read it
If we are going to have shoulds, then we can use it as specified in the RFC -- shoulds are compatibility points
Boris Motik: as we are using it, the meaning is exactly as in the RFC, so perhaps we should repeat it
I don't think we will be using other keywords
if you are departing from this default, you should advertize it clearly
the inventor is obliged to say what he really did there
Alan Ruttenberg: the only question I have is that the manuals say how to do this and make it typographically visible - is there any reason we shouldn't do that
in the w3c manual, it says explicitly how to use them
Bijan Parsia: this idea that we should have some kind of --- from vendors we should talk about some notion of conformance, and that we could ask that warnings be given in some form or another
we havent done any of that yet
Alan Ruttenberg: if you could put an issue in for this, it's distinct from what we're discussing and useful
Boris Motik: if people really depart from these things, it has to be clear that an implementation is really departing from the "should"
it would be useful for the implementer to say what part of the shoulds they did not implement; if a vendor says they are compliant, they should say that they are OWL 2 compliant BUT ...
Bijan Parsia: among our options are conformance labels, warnings ... and we can choose what we say about these, similar things occur in other W3C specifications - we can say that in order to conform with OWL 2 you must adhere to the shoulds ...
Boris Motik: I agree that the last thing you said is just a conformance label and I've put this into the spec
Alan Ruttenberg: the issue is if we are going to use "SHOULD" then we should follow the advice of the TR with respect to how we use them; if we are going to talk about conformance levels, that's an interesting and separate issue that we should put in and take up at another meeting
we also need to cite it as a reference and do the other things they say we need to do
Action on Bijan to write up this point of view
It would be good to have one of the W3c guys write up something on how to do this using XSLT
I want to get someone to commit to the writing
Alan Ruttenberg: Ivan would you write up the first draft?
Ivan Herman: ok
on irc or the wiki
ACTION: bijan write 1/2 of GRDDL pro/con document for presentation and vote in next f2f
ACTION: ivan write 1/2 of GRDDL pro/con document for presentation and vote in next f2f, Sandro to own it at F2F
Bijan Parsia: I'm confused about what I should be writing up on GRDDL - need a little context, but it came from the discussion two weeks ago
Sandro Hawke: the process could be that Bijan writes it, and everyone screams, or Alan writes it and everyone screams -- Sandro will provide the context to Bijan with regard to how to resolve this
from the minutes two weeks ago
Alan Ruttenberg: we were going to have a formal vote on this ...
Bijan Parsia: so where does that leave us?
Alan Ruttenberg: I would rather have Bijan contribute the piece he needs to create, and then have Ivan do the same with the other side
Michael Schneider: the issue is for OWL Full in the current semantics, the complement is relative to the whole domain, and the problem is that all datatypes, or all subsets of RDFSLiteral ...
the currently used URI for this is overloaded
there are two domains for this in OWL DL but only one domain in OWL Full
Alan Ruttenberg: originally, when we talked about complement of datarange, when we talked about the complement of 5 integer, you got all other integers ... but then Boris said that the complement would include all other datatypes, not just integers
Michael Schneider: what i propose is to just give it a strict name for this other complement
this is different from Issue 124 -- it is much easier to fix, all we need to do is provide an owl data complement uri
this came up during the discussion of Issue 124
Alan Ruttenberg: my thought is to include this in the same issue, rather than opening another issue
Peter Patel-Schneider: Michaels solution for this is perfect
Alan Ruttenberg: Michael, would you come up with a solution of this for our agenda for next week?
Michael Schneider: yes, I'll do that
Ivan Herman: I have written up some of the discussion I had yesterday with Bijan, but I am not as pessimistic about this as you are. I tried to write down what the choice is and next week we can vote on this and people can choose between the two options. It's not that big of a deal
Alan Ruttenberg: that's fine with me unless anyone has anything else to add
Additional Other Business
Alan Ruttenberg: we have a mail from a user asking about horn shiq and about why it's not in owl ... how should we respond
Bijan Parsia: is there any reason not to include it as one more profile?
depending on whether you count 3 or 5 or 7, depending on how you count the full versions of the little ones ...
Bijan Parsia: hornshiq is a distinct and interesting profile, so would this open the floodgates? there is a user who wants this ...
Carsten Lutz: if we want to consider adding this, is it interesting enough to become rec? that's not a very strong point for adding it - we already have one data complexity profile, so I'm not really convinced
Bijan Parsia: the other thing is that there are modeling problems that fall into hornSHIQ that are not relevant to the other data complexity profile - that's where he's coming from
otherwise i agree with you in general
but Christian was coming from both a modeling and performance perspective
Alan Ruttenberg: I wonder if we should have the requirements people capture the modeling issue
Carsten Lutz: I disagree, because that something was good for modeling is not a good reason to include something - there should be an additional virtue that it has when you don't use DL full
Alan Ruttenberg: I meant modeling in the sense of the inference you could make from it -- and whether the use case was compelling enough and the performance gain compelling enough to consider
so - we have on one hand some people are saying well use OWL DL, we have too many fragments already and it isn't sufficiently compelling; the alternative would be to say we will investigate a little more and you shoudl talk with our requirements people about it
Alan Ruttenberg: I like what Bijan says -
Ivan Herman: I think Alan should respond
ACTION: Alan to respond to the email along the lines Bijan suggests above
Alan Ruttenberg: last item -- Bijan brought up the issue of accessibility guidelines, and the work that needs to be done to follow those guidelines in our documents
what work needs to be done, how do we get started
Sandro Hawke: I know there is something to be done but don't know how much work it is
Bijan Parsia: it's certainly the case that for our images we need to have alternate text, tables are often hard for assisted technology without additional mark-up
there may be work to be done to make sure that the tables are good enough
there are some tools that check from an accessibility point of view
Ivan Herman: I don't know how extensively we use java scripting - that would require some additional explanation in the text
Alan Ruttenberg: it seems like we need some research on this, and someone to review our documents ... maybe we can discuss on the chairs list how we can get additional information and get back to the group with some harder facts
Alan Ruttenberg: AOB?
ACTION: Alan to confer with chairs list about how to get more information about what we need to do re: accessibility