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

At 2:50 PM +0000 1/12/07, Uli Sattler wrote:
On 12 Jan 2007, at 14:37, Jim Hendler wrote:



[...]

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.    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.  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.   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.


Jim, I am afraid that -- while we can indeed have measures for a 
fragment's tractability/complexity --  we don't have such a measure 
for a fragment's "usability". I believe that a fragment's usability 
strongly depends on the domain it is used to model and on the 
background of/taste of/tools used by  the person modelling it. A 
similar observation holds for "how difficult is it to explain 
fragment X to a user"? In my understanding, e.g., people used to 
believe that qualified cardinality constraints were too complicated 
to explain, and therefor not included in OWL -- yet their absence 
caused modelers a lot of pain and trouble.

Cheers, Uli


Uli - this is what I'm afraid of.  It turns out that most standards 
do indeed try to consider these things - the real world in fact cares 
a lot about these things - and some are easy - for example, adding 
some minimal cardinality to OWL fragments doesn't add theoretical 
complexity, but in RDF/XML syntax, it requires learning the 
"restriction" syntax, which is more complex than just making 
assertions about classes and prperties...

but let me be clear - I think there are two approaches that would be 
valid for the Working Group - either take usability and the real 
world into account, or leave the task of defining other OWL subsets 
to people who do.  What is a mistake is for the group to take it on 
and do it on purely theoretical grounds - all we'll end up with 
another travesty like OWL Lite

if the WG wants my opinion, they should remove that topic from the 
scope - but if that's not viaable, then expect that there will be 
those of us who insist that the WG pay attention to things that are 
not "measurable" and require us trusting people's instincts and 
experiences, which makes many formalists very nervous.
  -JH

-- 
Prof James Hendler				hendler@cs.rpi.edu
Tetherless World Constellation Chair		http://www.cs.umd.edu/~hendler
Computer Science Dept			301-405-2696 (work)
Rensselaer Polytechnic Inst			301-405-6707 (Fax)
Troy, NY 12180