See also: IRC log
jar, i've just emailed amy, but we better call her for help getting a bridge line
<jar> talking to ralph now
thanks
<jar> 'one moment'
<jar> zakim. this will be awwsw
<jar> dialing in now
jar: (Gives tim an update on discussion for the past few weeks.)
timbl: Lots of people see a diff btwn IR and representation over HTTP.
jar: Alan's view seems to be a sequence of bits w mime type, varying over time.
<jar> alan would like to rule out content negotiation
tim: we can argue over whehter IR fundamentally *is* a mapping or *has* a mapping, but that doesn't answer how the web works. To me writing rules like david has been writing, that's valuable, because it can be used in software.
jar: It's one thing to say what
triples tabulator should assert, and much harder to say what is
an IR, but I'm trying to look in the middle: when is an RDF
graph consistent w server behavior?
... e.g., you've asserted that 3 is not an IR. So if I have a
URI that denotes 3, we want that to be indicated as disjoint
from IR.
... So we want to know not only what is an IR but what is *not*
an IR.
... But the AWWW doesn't give enough guidance to know whwther 3
could be an IR.
<jar> timbl: 3 is not an IR is more a statement of programming language types
timbl: We're not doing physics,
we're doing engineering. One could design a system in which 3
is an IR. You can write a programming language in which numbers
are strings ... it's a question of design.
... Could you consider 3 to be a document in which the value is
3? You could do that, and it's not fundamentally inconsistent,
but it's more complicated to program, more difficult to think
about, etc.
jar: who makes those decisions?
You, me, Xiaou Shu?
... Those decisions need to be articulated in a waht that lets
me know what to do.
<jar> timbl: an ontology should be grounded in code
timbl: We're going in the right
direction by writing an ont and rules.
... But I don't expect to complete the sentence: "An IR is a
____"
jar: Yes, you want things to be grounded in code, but the code doesn't know what to do with things it hasn't seen before. E.g., is pamphlet an IR?
timbl: That's why we generate an ont.
jar: We cannot enumerate every
possible thing that isn't an IR.
... We don
... We don't want an engineer down the road to be surprised by
timbl saying "3 is not an IR".
timbl: Why do you need to
enumerate them all? We build a system, and people connect it
up.
... and we say "IR is not a physical object", but we don't have
to provide something that will answer every question involving
IR. That would take an infinite amount of time.
jar: someone trying to connect another ont to this system, it seems useful to have some sort of guidance.
timbl: i think the second gen of
people using SW tech will not argue about it. They'll know what
would happen if they did -- they've internalized it.
... When you first introduce OO programming it is strange, but
the next time it comes easily. people won't keep going back to
fundamentals.
... Because there will be such a mesh of people with such a
large amount of understanding about it that they can ask around
the corner.
... You should write something down, but you should have
limited expectations about that. If we write it for a general
audience and then someone comes back saying "that doesn't help
me", and we would have to write something different for that
person, or sit them down with lots of beers,e tc.
jar: there are various entities that i'll have to deal with, and there may be millions of edge cases. And without clearer guidance I'll have to do 303s which seems unfortunate.
timbl: If you're building an ont of works and concepts, there may be lots of ways you can go, you're doing engineering not physics. Deciding whehter something is an IR is deciding whether you want someone to be able to read it.
jar: But it's a decision I'm not
free to make, because when i think something is an IR, you come
and say it is not.
... Suppose i want to give a URI to the meaning of a fragement
of a doc, or for a representation. Things that are arbitrarily
close to the border between IR and non-IR.
timbl: If you take apart the software system, most of the things a C program talks about are not files.
jar: People like to do URIs with 200 responses for all sorts of things, adn I cannot tell ahead of time whether they are IRs.
timbl: If you want to model a blog, you'll have a much more fine grained ont than the average person.
<jar> timbl: if you want to model a blog, you have a more fine grained ontology than the ordinary person
timbl: But if you ask serious bloggers, they'll say the blog is the whole conceptual thing.
jar: But i'm asking what i can use the URI for?
<jar> dbooth: but a journal article is close to an IR too
<jar> timbl: if you go to the handle system, the 200 you get back is not always the article
timbl: journal article is an IR, it's got meaning
<Stuart> Try: http://dx.doi.org/10.1002/cpe.1238
timbl: but if you go to the publisher you'll likely get a 303 because they want to treat it as an abstract object.
<jar> timbl: a journal article is inherently an IR
<timbl> http://example.com/fred/myarticleinccvm.ps
dbooth: so it sounds like you're saying that a journal article can *also* be something more than an IR.
<timbl> http://publisher.example.com/journals/cacm/234/45
<jar> timbl: the abovenamed thing may be the same work as ...
<Stuart> Try a concrete (DOI) example... http://dx.doi.org/10.1002/cpe.1238
dbooth: So an IR is permitted to have *some* additional properties, but not others.
timbl: On this call i want to
write an ont, not prose about what IR is.
... e.g., IR is disjoint with RDF literals.
... Because there is no single upper ont that everyone has
bought into, then the first thing we'll discuss is whether an
IR can be a sign, then a huge discussion will ensue about what
we mean by "sign".
... So I'm wary about simply giving a definition of IR. I don't
think we can give a decision algorithm.
jar: Alan recently brought up a nasty issue about URIs.
timbl: The TAG wrote about that earlier. We called it a URI ladder.
Alan's issue on URIs: http://lists.w3.org/Archives/Public/public-awwsw/2008Jun/0019.html
<Stuart> http://www.ietf.org/rfc/rfc3986.txt section 6, 6.1 and 6.2
dbooth: The problem i see with trying to prohibit a resource from being both an IR and a Person is that that is architecturally no different than trying to prohibit a resource AKT from being both AKT1 and AKT2, i.e., the resource can simultaneously have properties of both AKT1 and AKT2 or both IR and Person.
timbl: But the ont can specify IR such that it is disjoint from physical things.
dbooth: Yes it can, but what is the architectural justification for excluding some things and not others?
[ ran out of time ]
ADJOURNED
This is scribe.perl Revision: 1.133 of Date: 2008/01/18 18:48:51 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) No ScribeNick specified. Guessing ScribeNick: dbooth Inferring Scribes: dbooth Default Present: DBooth, +1.617.253.aaaa, TimBL, jar, stuart Present: TimBL JonathanRees DBooth StuartWilliams Regrets: Noah Got date from IRC log name: 01 Jul 2008 Guessing minutes URL: http://www.w3.org/2008/07/01-awwsw-minutes.html People with action items:[End of scribe.perl diagnostic output]