comments on the owl 2 last call


a little belated but maybe still of use (if of use at all). i'm coming  
from a web designer background. my knowledge about description logics  
is mostly self taught and my interest in semantic web technologies  
stems from my obsession with multifaceted structuring of information.  
that said, let's comment:

i have to admit that i'm quite perplexed that this datatype is at  
risk. how can OWL 2 want to be 'on the web' but not support it's  
single most important datatype? how can it be so difficult to  
implement a datatype, beside all the real complex problems involved  
with logics, reasoning etc? i strongly suggest that implementors that  
want to support OWL 2 without XMLLiteral may (of course) do so but  
have to make a note to their users (and can't claim full OWL 2  
compliance). instead of laying the burden on the users to first check  
if a given software supports XMLliteral. i don't say that every OWL 2  
application needs XMLLiteral. but if it's on the web i want to be able  
to rely on that.

restrictions versus materialization:
this problem is tricky and it needs more elaborate discussion at least  
in the primer: it's own chapter, explicit mentioning of the concept of  
"materialization", the implicatons of rdfs:domain and rdfs:range, how  
your restriction can be my materialization and vice versa, in what  
ways owl:restriction can be used and in what ways it shouldn't etc. i  
think a lot of misconceptions about OWL are connected to this topic  
and it should be discussed exhaustively.

the introduction to the primer states that it will get further work  
which is good. it has muched improved in the recent version but it  
still needs much more examples. i would prefer if the primer didn't  
try so much to use every OWL element once but if it was more of a  
general guide to ways of modelling common situations, giving example  
constructs etc. for example i asked a question on public-owl-dev last  
week and alan rutenberg was so kind to propose a solution. it took me  
quite a while to understand that solution (and also to a colleague  
with a very strong semantic web background it was not obvious). while  
i could understand it after some pondering there is no way i could  
have come up with it myself in the forseeable future. i need examples,  
to copy!

more introduction:
i would suggest that the introduction of every element in examples  
gets its own document. the new (and useful) document overview names 4  
parts as "for user" but those 4 parts merely give an overview. then  
you're stuck with the core specification which for non logicians is  
shockingly unreadable. there is a huge gap between the two which needs  
to be filled.

it doesn't help either that there are all sorts of syntactic  
variations. why can't we stay with infix notation? a trained logician  
should have a much easier time switching from prefix to infix then the  
ordinary web developer has the other way round. if OWL is meant to  
introduce logic into the web data on a large scale then please: climb  
down from the hill;-) and why do properties have different names in  
different syntaxes? why does manchester impose even more new  
constructs? and there is a serious lack of publicly available  
converters, as web services and as software (but at least those  
available should be mentioned in the primer).
and where have the pictures gone? painting down the graph is sometimes  
my last resort when i'm trying to understand what's going on (eg when  
alan gives e an example ;-) and can be really helpful. it's the most  
easily understandable syntax. manchester may be easy and terse to  
write (maybe it's the desperate sem-hackers tool of choice?) (well,  
maybe i'm deformed by to much triple talk...)

why is there a new reification syntax? why not use the one that RDF  
provides? i can't spot a difference neither semantically (that may be  
me, of course) nor in terseness or syntactic elegance. it's the same  
triple bloat, just with new element names, isn't it?

OWL 2 namespace:
why isn't there one? wouldn't you, when you introduce new element and  
sometimes even new rules, refer the user/parser/reasoning engine to a  
new version of the spec? how does this deal with OWL softwares and  
services that don't get updated to OWL 2? does it degrade gracefully?  
degrading gracefully is quite important on the web.

thanks especially for punning! and thanks for all the other work too!  
indeed i do (soemhow) see what an enormous amount of work these  
documents represent and i'm very thankful for it, despite my rant above!

thomas lörtsch

Received on Friday, 15 May 2009 18:28:58 UTC