W3C home > Mailing lists > Public > public-owl-dev@w3.org > January to March 2007

New draft charter (was Re: Responses to "Draft of charter for NextWebOnt (Proposed) Working Group")

From: Bijan Parsia <bparsia@cs.man.ac.uk>
Date: Tue, 16 Jan 2007 10:41:39 +0000
Message-Id: <F5252251-38CE-4969-A317-CF88700A4FCF@cs.man.ac.uk>
Cc: public-owl-dev@w3.org
To: Jim Hendler <hendler@cs.umd.edu>
There are some interesting debates on substance in this thread.  
However, I think it distracts a bit from the discussion of the  
charter *per se*. So, I would ask, for clarity of discussion, that  
people split off substantial substantive debates into separate  
thread. Here I'm trying to respond as much as possible wrt to the  
charter alone. I've made some changes to the draft to address the  
concerns Jim has raised.

On 12 Jan 2007, at 14:37, Jim Hendler wrote:

> As I promised - I might have some more comments - here are some:
[snip]

> I agree with many of Bijan's motivations (seeAlso his blog) but I  
> have some concerns as to whether this fully considers some issues  
> about the usability/marketability of OWL:

As anyone who reads the comments or knows me, I (and others) have put  
an *enormous* amout of consideration into various issues about the  
usability and marketability of OWL. Of course, not everyone will  
agree with my judgments on these matters, and for some I am reluctant  
to come to settled judgements due to lack of solid information, and  
variations in situations and requirements.

Any charter builds in *some* assumptions about what's usable, useful,  
and saleable while trying to balance in other considerations as well.  
Part of the consensus building and standardization process is making  
sure that enough of everyone's perceived interests are addressed  
sufficiently to make it worth all our while. Since not every  
stakeholder...including future WG members who, after all, will be  
putting up substantial resources to complete the work...is present  
for this (or any) charter discussion, I'm inclined to be very careful  
about what we build into the charter. I want there to be the  
flexibility for the group to meet its members' (and other  
stakeholders') needs; I want support for "getting it done" on  
schedule; and I want the probability of good outcomes. I'm pretty  
sure that the current charter supports these desires on balance.

> Anyway, my comments on this are that I think these are good goals,  
> but when we look at the deliverables I don't see some things I  
> think are needed
>
>  1 -  There are a set of existing recommendation docs, esp. the  
> model theory, the guide, the reference manual and the overview  
> which are important to OWL.'s use.   The new charter says the group  
> will:
>
> The working group will work to ensure a smooth transition from OWL to
> OWL 1.1 by providing suitable outreach documents (whether new or as
> updates to existing documents), and by striving to maximize backwards
> compatibility, especially of ontologies.
>
> I'd like to see a specific commitment to extending the 4 documents  
> I mention above - doesn't mean new ones couldn't be written, but a  
> group updating a spec should update the mandatory documents

On the grounds above I'm very reluctant to see this this into the  
charter. For example, I think that refactoring the Semantics and  
Abstract Syntax document into separate Functional Syntax, Semantics,  
and Mapping to RDF documents is likely a very good thing (all these  
areas are in the current S&AS document!). I am *very* reluctant  
*either* way to constrain the working group in this regard. That is,  
I am comfortable to leaving it to the group whether to directly  
update the S&AS document "in place" or to split it up. I am not  
comfortable seeing either as part of the charter.

It is not unusual for a WG to change the "shape" of recommendations  
in a new incarnation (see, RDF Model and Syntax vs. the set of spec  
RDF Core developed), so I don't see allowing that freedom is out of  
the norm for W3C charters. OTOH, some charters explicitly require  
evolution:
	<http://www.w3.org/2003/09/xmlap/xml-schema-wg-charter.html>

I believe that there are decent arguments on both sides (for various  
cases) and that the WG should marshall the evidence they need to make  
a decision (e.g., comparing the documents; thinking about what needs  
to be added; looking at current WG practice; checking log files for  
traffic; etc. etc.).

> 2 - I am most worried, personally, about the "example, identifying  
> useful sub-languages that are (more) tractable and/or efficiently  
> implementable, e.g., with standard relational and deductive  
> database technology" statement." (from [1])
>  One mistake we made in the original OWL group was we did not spend  
> enough time thinking about "usability" instead of theoretical aspects.

This is a tendentious history.

>     While I have no problem with the tractible subsets document, it  
> very much does NOT answer the mail on this.   For example,  
> syntactic considerations, ease of use considerations, and usability  
> concerns are not included in that document - it is too much a  
> theoretical aspect.

I think you are conflating two things: 1) the design considerations  
for subspecies and 2) the presentation in the tractable fragments  
document. At least two of the fragments, EL++ and DL Lite are  
languages that take a panoply of user and application issues into  
consideration. Effective tractability is one important user  
consideration (EL++ balances the high expressive needs that many life  
sciences ontologies require with a proof procedure that has been  
shown to scale to the large size of many current life sciences  
ontologies). DL Lite captures a large number of UML class diagrams  
*and* can be realized *on top* of existing relational database  
technology, thus can be used for database integration, even "on the  
fly" database integration (and you can use restricted UML to directly  
build DL Lite ontologies). These considerations are mentioned in the  
document, though not highlighted. They are more extensively discussed  
in the referenced papers.

>   Oracle, for example, announced at ISWC the subset of OWL they  
> intended to support, and it was not a maximal tractible subset - it  
> was a smaller subset that they chose for issues including how  
> straightforward it was to explain to their users.

Teachability is a consideration, though a tricky one, I think (after  
all, it varies enormously with the user base). And only one of many  
considerations.

The TF document is not a proposal for a set of species, it simply  
highlights "large" fragments with "interesting" properties. This  
should help people understand what features "combine" well, and so  
can help them develop the fragments they think hit the right sweet  
spot (one problem with OWL Lite is that the features "stripped"  
didn't actually lower the complexity *or* allow for substatively  
distinct implementation techniques). Identifying which species, if  
any, should actually be identified is exactly the sort of things the  
WG should hash out.

One thing to recognize about the TF doc is that the fragments chosen  
aren't *arbitrary* (or solely motivated by formal complexity  
results)...there are many many theoretically interesting fragments  
(from many perspectives, including that of formal worst case  
tractability of key inference services....consider propositoinal  
logic! (which does have a significant and interesting user base)).  
The five languages selected are largely *representative* of distinct  
families of languages that have interestingly distinct strengths and  
weaknesses expressively, implementationally, and support for various  
applications. They all have *champions* and *implementations* and  
*natural user bases*. These are, to my mind, key (if not necessarily  
sufficient) components to a good profile/species/etc.

I hope this lays to rest the concern that "merely theoretical"  
considerations are driving any aspect of OWL 1.1. For many of us,  
*effective* tractability is a constraint on a useful and useable  
species. Thus, the TF doc provides a framework for discussion, not a  
presentation of all the issues that go into designing a species. To  
make this crystal clear, I added "usable" to the relevant section of  
the charter. I do not at all regard this as a substantial change, but  
it might help clarify:

"""Rationalization of the species of OWL. For example, identifying  
useful and usable sub-languages that are (more) tractable and/or  
efficiently implementable, e.g., with standard relational and  
deductive database technology."""

Secondly, I tweaked the deliverable to be "Tractable Fragment/ 
Rational Species (RaS)" to highlight the options available to the group.

>   I think it important to have a named subset of OWL that falls in  
> this area - if the group is not willing to take this on, then the  
> "rationalization" goal should not be included.
>
> 3 - Whil I like the idea of refinement of OWL specification, I  
> would have trouble supporting a recommended normative syntax that  
> is a "Deterministic and round-trippable" mapping from the syntax to  
> RDF graphs - as this means arbitrary RDF graphs expressing OWL  
> relations would not necessarily be able to "play"

That doesn't follow. Deterministic and roundtrippable means that for  
every Functional Syntax document there is one, and only one, RDF  
graph (by this mapping) *and vice versa* (assuming that you otherwise  
match the species). There maybe "trivially" equivalent graphs that  
map into distinct FS documents...but that's ok! They'll still be  
equivalent, but their syntactic distinctiveness will be preserved.

In other words, this is not a constraint on either syntax, but on the  
mapping. There could be other mappings (including non-deterministic  
ones), just as there can be grammars which require backtracking for a  
language that has an LL(1) grammar. Both the nasty grammar (which  
might be more readable) and the nice LL(1) grammar will, in the  
appropriate setting, determine the same set of strings, but one has  
computational, implementational, and usability burdens that are worth  
avoiding. So too with a non-deterministic vs. a deterministic  
mapping. The former is somewhat easier to specify, but is nastier to  
work with overall.

So I don't see a need for a charter change to address this. If such a  
mapping proves infeasible or impossible, the working group can drop  
the extra constraints and go for a nondeterministic, non- 
roundtrippable mapping. (Remember, dropping functionality is easier,  
under this charter, than adding.)

[snip]

> 4 - the scope seems awfully ambitious for a 1 year group, esp. when  
> you consider the above.  I would recommend deciding whether to take  
> all these things on and allowing more time, or cutting out some of  
> these  (do XML spec or Owl species or refinement of spec - not all  
> of them)

The charter allows for the dropping of tasks or functionality in  
order to meet the schedule. Since a good deal of the substantive work  
is done already (including some implementations!!!!) I don't think  
it's pie-in-the-sky ambitious. The charter allows for the WG to  
"adjust down" on the fly if they need to.

Also, there are several months before the conceivable start of the  
working group. A good number of people will be working with the OWL  
1.1 language and specs (and on them) during that time. For example, I  
plan to do a big chunk of work on the XML syntax and on the RDF  
mapping, and their attendant documents. Furthermore, it was suggested  
at OWLED2006 that interested parties could start friendly telecons  
*before* the working group started. We already have gotten lots of  
feedback on the design, and I expect more. So, we'll have pretty well  
polished stuff and a good deal of momentum going in.

> Finally - a quick note - please notice that compared to my earlier  
> reservations about this WG, I feel this charter is a major  
> improvement and these are all constructive criticisms based on (too  
> much) experience with WGs, not meant to be "show stoppers" - but  
> remember that for this WG to work, there must be players from  
> outside academe, and for these more applied players, the  
> sturcturing of the WG qua WG is as (and sometimes more) important  
> than the list of the deliverables.   Getting the charter right is  
> really important.

For the record, many of the people giving input (e.g., at OWL ED,  
etc.) are not academics, and the submitters and "supporters", all of  
whom have expressed willingness to participate in a working group,  
include several from outside academia. We tried to model the charter  
on WS-Policy's (a WG that was mostly non-academic), and considered  
several other current charters while developing this. This is not to  
say that it can't be further improved, but that such considerations  
are already in play. Regardless of in or out of academia, the charter  
needs to define an attractive and effective working group. That is  
the overall guiding principle for me.

The refined proposed charter draft is at:
	http://homepages.cs.manchester.ac.uk/%7Ebparsia/2007/temp/ 
15-01-2006NextOntProposedCharter.html
and attached.

Oh, I should add, again, that this proposed charter draft has no  
official status.

Cheers,
Bijan.



Received on Tuesday, 16 January 2007 10:41:25 GMT

This archive was generated by hypermail 2.3.1 : Wednesday, 27 March 2013 09:32:54 GMT