Re: Working on glossary

>Since I contributed last, Pat and Graham have been going at it
>hammer-and-tongs
>
>See:
>GK: http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0127.html
>PH: http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0130.html
>GK: http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0131.html
>PH: http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0132.html
>GK: http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0061.html
>
>
>I am grateful for the attention paid to this question.
>
>But, chaps, what I wanted to make was a glossary for the layperson which
>would help their initial understanding of some of our terms. I see,
>again, how dubious some of these terms are.
>
>Pat knows, from an exchange which came to nothing probably because I
>didn't express myself that well, that I am concerned about what I call
>the `temporal coherence' of stuff `on the Web'. This does seem to be an
>issue and, I can see, it's an issue which will not disappear just by
>defining (and redefining) the words we use. We have to offer to the
>newcomer a definition which allows for temporal variance without
>completely denying the usefulness and significance of the words.

Well, if there is an issue (which is indeed important and all, but) 
which we have decided to shelve, then we ought to say that clearly, 
not give definitions which give the people what they want to hear, 
but don't in fact correspond to what we are saying.  RDF as it 
stands, seems to me, *doesn't* deal with temporal variance. Maybe it 
should, but it doesn't. (It doesn't deal with lots of other things, 
either, of course.)

>For those who haven't been following the thread, much of what has flowed
>past has been about the validity of defining a resource at all given its
>changing nature (whether incidentally in a changing world or
>functionally as a news service, for instance). Graham initially wanted
>to point out how separated the ideas of entity and resource were as in,
>for example, `today's weather forecast'. Then the discussion started...
>
>Just to quote some of GK's and PH's dialogue:
>
>Graham refers me (usefully, thanks) to
>http://www.w3.org/DesignIssues/Fragment.html where a distinction is made
>between a fragmentid-less URI which refers to a generic document and a
>fragmentided URI which refers to an identification within that document
>(and not just, as we know, a point in that document :-). Thus no '#'
>suggests just generic access and a '#' suggests an identification (eg,
>of a resource). Then he says:
>
>GK: `I find it difficult to unify the RDF idea of a resource (a thing
>that is identified) with the general Web idea of a resource (a thing
>that is queried or accessed) in light of these comments.'
>
>Me: So do I, Graham. So do I.

Pat: So do I; but I thought that this particular discussion had been 
already dealt with, and that the broader terminology was now 
accepted. If not, then lets go back and get it sorted out as quickly 
as possible.  But if so, then let the glossary reflect the meanings 
we are using, not the ones that people wish we were using.

>I also find it difficult to generalize
>this distinction to non-document resources (you know the list, people,
>organisations, trees in the park, numbers,...)
>
>GK: (on the time-variance track) `The approach that I have seen adopted
>is to treat such things as sequences of values, or time-varying
>functions.' ... `Then, the resource can correspond to the current
>weather forecast, but its extension includes the set of all weather
>forecasts for all times;  the particular member of that extension one
>retrieves depends on when the retrieval is performed.'
>
>Me: Yes.

Pat: But it isn't this easy. Getting this right is tricky. For 
example, if the temporal extension of <uri1> is time sequence A, and 
of <uri2> is time sequence B, then presumably the triple <uri1> foo 
<uri2> is not supposed to say that foo holds between the A and B, but 
that a sequence of foo-relations holds between the elements of the 
sequences. But *which* elements? Always those at the same times? Or 
do we allow relations like Yesterday which are true between <uri1> 
and <uri1>-yesterday? What if foo holds between A and B only on 
weekends, say? And so on.

The use of words like 'current' sounds fine to a lay reader but it 
gives any logician nightmares, since its an indexical: it means 
different things at different times, like "now". Think about it. 
Suppose we actually had a logic with times, all sorted out with 
time-orderings defined and so on. Now, what time does a word like 
"now" denote? Answer: whatever time it was when the word was used. 
But what does it mean to 'use' a piece of RDF? RDF is put on websites 
and left there; it isn't speech, said by one person to another at a 
time and place. When was the RDF "used"? When it was composed? Or 
posted? Or read? Or used in an inference?  Indexicals literally make 
no sense in an assertional logic.

>This is why I think it's important to say that a `resource' is
>an entity identified by need (if you like),

No, I don't like that at all. In fact I don't know what it means. 
(Identified by need???)

>  that the act of identifying
>it as a resource includes an understanding of such characteristics. This
>may not be its most precise definition but it would suggest for a new
>reader the role of the concept.

It would suggest a whole lot of stuff that we wouldn't have a hope in 
hell of ever getting clear, let alone formalized in any kind of 
useable inference language.

>PH: `As I understand it, the real-world objects *are* resources, so the
>idea of being retrievable on the web simply isn't applicable to
>resources in general.'
>
>Me: Retrievability is a characteristic of some resources, of course, but
>is not the crucial aspect of their identification even if the identity
>of a resource (`today's weather forecast') is expressed in terms of its
>retrieval. The resource is identified as `today's weather forecast' and
>its usefulness (the `need' for identification) is in its timeliness BUT
>the act of retrieval is not what makes it a resource.

OK, but I really do not follow this talk of something being 'made' a 
resource. As far as I can see, in the sense we are using the word, 
everything already is a resource. Being a resource is just a way of 
saying that it exists (in fact, according to Brian, not even that.)

BTW, I also don't follow this notion of 'identification'. What would 
it mean for something to *not* have an identification? There is a 
similar notion of 'having an identity' that occurs in some W3C 
documents, which has a nice philosophical ring to it but which seems 
to be meaningless, since in the philosophical sense *everything* has 
an 'identity'; that's just a way of saying it is identical to itself.

>
>GK: `The alternative view I'm trying to offer here is that real-world
>objects are not resources, per se.  In addition to the octet-sequences
>that are web-retrievable entities, I'm suggesting that the real world
>objects are
>(also) part of the resource extension I mentioned previously, rather
>than the resource directly.'
>
>GK: `... but such (time-sensitive) semantics would be part of the
>semantics of a resource, about which the RDF model theory you have
>described seems to be agnostic.'
>
>Me: Ah-ha! Yes. This is how I think of it. The timeliness of information
>belongs in the semantic of that information but should not, in some
>sense, invalidate the identification of it as a resource. Again, a
>resource is an entity identified as required: the semantics may suggest
>that the entity will change.

I cannot follow this at all. Can you apply it to an example? Take 
'the current weather forecast', say, and assume for simplicity that 
it is one of Rain, Dry or Cold, and is issued every day. Now 
something like 'today's weather' is presumably one of Rain, Dry or 
Cold on any particular day, right? That's what makes it 'timely'. If 
it were the entire time-sequence <....Dry, Rain, Rain, Cold, Dry, 
....> it wouldn't be timely. But on this proposal, the Uri 
<todays-weather> denotes the entire sequence, so if we use it as the 
subject in a piece of RDF we are saying that the entire sequence has 
some property, not that todays' weather is suitable for playing 
cricket.

>
>PH: `In the meantime however we could also just put this issue off to
>the future, and the current MT be thought of as kind of instantaneous
>time-slice of this extended temporal semantics.'
>
>Me: Oh, please, no. Or at least, let me get my glossary done with a nice
>clear message for the newcomer while we perhaps agonize about the real
>meanings. I don't think we have forced ourselves to consider that RDF
>refers to `a time-slice of an extended temporal semantic'...

I think that we have. (As far as the MT is concerned we have, 
anyway.) So the glossary ought to reflect the reality of what we have 
done, not our dreams of what we ought to have done; still less the 
novice's fantasies about what we ought to have done, no matter how 
clear they might be.

>
>The ref
>(http://www.ietf.org/internet-drafts/draft-masinter-dated-uri-00.txt)
>that Graham includes is a good one, I admit. The idea of a duri: URN
>protocol to state that `I am referring to the page as it was then' is,
>of course, a valid one but it does not fit in with the requirements of a
>practical Semantic Web.

I disagree: I don't think the SW can manage *without* something like 
this, in fact, in the long (or even medium) run. The SW isn't for 
PEOPLE to use, remember, its for MACHINES to use. People can use the 
ordinary web. But if some B2B software agent is trying to figure out 
when the money was transferred to the account, I would like for it to 
be *very* clear about things like the difference between today and 
yesterday.

>Larry has it right but it's not going to help.
>It's great for putting in legal documents. However, the Semantic Web is
>going to have to cope with time-variant data and not just deliberately
>time-variant (like today's weather) or with time-variance created by
>retrieval or time-variance due to decay.
>
>Help. What shall we tell the novice?

Just tell them the truth. If its not what they want to hear, then its 
time they were told the way the things really are.

>  `Resource' (and its relationship to
>`entity') must be in the vocabulary of anyone starting to read the RDF
>documentation.

Actually, this usage of 'resource' is unknown outside the W3C 
community. Unlike words like 'entity' and 'thing', it isn't going to 
have a lot of surplus meaning-baggage: its a technical term we can 
define any way we like, within reason.

Pat Hayes


-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Wednesday, 12 September 2001 19:01:07 UTC