W3C home > Mailing lists > Public > www-tag@w3.org > September 2007

Re: New Editors Draft of the httpRange-14 Finding

From: Alan Ruttenberg <alanruttenberg@gmail.com>
Date: Sun, 30 Sep 2007 01:24:55 -0400
Message-Id: <C5B052D9-7E11-4397-B3F7-8582C60E5FC2@gmail.com>
Cc: "'www-tag'" <www-tag@w3.org>, "'Jonathan Rees'" <jar@creativecommons.org>
To: Rhys Lewis <rhys@volantis.com>

Hi Rhys,

Been busy - but better late than never.

On Aug 29, 2007, at 3:53 AM, Rhys Lewis wrote:

> Hello Alan,
>> -----Original Message-----
>> From: Alan Ruttenberg [mailto:alanruttenberg@gmail.com]
>> On Aug 24, 2007, at 3:53 AM, Rhys Lewis wrote:
>>> Hello Alan,
>>> My impression is that the language in AWWW is deliberately that way
>>> because there is no hard and fast rule about what defines  
>>> 'equivalent'
>>> representations. I suppose that we might say that authors make
>>> assertions about equivalence when they create multiple  
>>> representations
>>> and offer them via content negotiation.
>> It's ok if they make the assertions. But there needs to be a  way  
>> to see that they make sense (at least if the plan is that  the SW  
>> is to be used for science).
> But that's the core of the problem isn't it? We can encourage  
> people to
> make assertions that make sense, but there is no way to check it.  
> At some
> level, languages like RDF and OWL help authors make consistent  
> assertions.
> I'm forever checking my ontologies to make sure I haven't written  
> complete
> rubbish. But I can still assert that the moon is made of cheese if  
> I wish.
> RDF lets me make such assertions with precision, but not necessarily
> accuracy.

Yes, but if you make the assertion that the moon is made of cheese,  
then we can do an experiment and determine whether you are right or  
not. That's because we know some things about cheese and the moon,  
and general science.

But the problem with "representation" is that lacking any definition  
that isn't circular. A person can't even figure out whether its  
correct or not, never mind a machine. So really, this isn't an issue  
of RDF or OWL or machine understanding and verification. I can't even  
get started on the machine stuff because I myself can't figure out  
whether someone has done this representation thing correctly or not.

Now although we can't check necessarily check whether our assertions  
are true, we can at least detect many cases where they are incorrect,  
because the combination of assertions we've made and collected can't  
possibly be true. There's a vast difference, BTW,  between doing work  
in  RDF as compared to OWL. In RDF it is impossible to look for   
errors except by staring, because it's impossible to have a machine  
detect an inconsistency. OWL is different. OWL can express enough  
that it *can* detect inconsistencies. I routinely find errors in my  
own ontologies and others by simply running a reasoner and seeing if  
there are unsatisfiable classes or inconsistencies. Adding a few more  
assertions that encode common sense in the domain of discourse  
flushes out more. You could have a look http://bio.freelogy.org/wiki/ 
Debugging_the_bug for an example of this, where the reasoner detected  
120 errors when merging a couple of databases, given only the  
simplest added assertions. As we add more you can be sure more errors  
will surface.

One of the reasons that I like working with OWL is exactly because my  
machine starts to be a help when I'm writing ontologies, instead of  
just presenting me stuff to proofread. As the amount of knowledge  
adds up it becomes impossible for a person to proofread as carefully  
as an OWL reasoner can.

> I don't see that web technology can provide any more guarantee of  
> correctness of assertions than a printing process that allows  
> papers to be distributed in learned scientific journals. Any sense  
> of correctness for papers needs to be applied above that level,  
> during the refereeing process, for example.

It can't detect correctness, but it can detect come kinds of  
incorrectness. And that is a huge help!

Regarding equivalence of representations. Fine, if you can't tell me  
when representations are equivalent, just give me *any* constraint on  
what can be in a representation. Anything at all that I (a person,  
not a machine) can use to audit someone's server behavior to verify  
that it makes sense.

Received on Sunday, 30 September 2007 05:25:14 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:56:18 UTC