Re: work in progress on dual-use theory

On Fri, Jan 21, 2011 at 3:56 PM, Nathan <nathan@webr3.org> wrote:
> Jonathan Rees wrote:
>>
>> So I'm making a model of the web in which all info-resources are
>> 'fixed resources' (Tim's term).  I figure that if we can't spin a solid
>> story around this, the whole enterprise is hopeless.  I think the
>> following
>> is pretty good, except for a gaping hole (inconsequential I think)
>> around exactly which resource is 'at' a URI. This goes back to our
>> 'phlogiston' discussion.
>>
>> Without 'phlogiston' the theory becomes immensely simpler, but TimBL
>> has imposed this as a requirement and shown little flexibility around it.
>>
>> Jonathan
>>
>>
>> Term: 'directive' (class)
>>  Def: Information supplied in addition to content that is
>>  intended to direct the interpretation of the content.  Content-type
>>  header or equivalent, content-language or equivalent, perhaps
>>  others.  (Expires: header??)
>>  Definition source: invention of JAR
>>  Similar to: metadatum, HTTP entity header
>
> is compression a factor?

I don't think the answer is important. You could answer either yes or
no and build a theory that works. I think I would deal with compression
in a way that is most useful to one or more applications (such as
Tabulator). It's probably most useful if you take HTTP decompression
to be already applied, but it would work just as well to take HTTP
compression as one of the directives.

>> Term: 'fixed resource' (class)
>>  Def: Content + directives, produced through some real process
>>  but with identity generic across physical incarnation.  I.e. it can
>>  be incarnated in multiple locations (disk drives, etc), each of
>>  which might be said to hold a copy of the fixed resource.
>>  Definition source: TimBL genont
>
> assume fixed resource doesn't have any properties which vary over time?

Please read Tim's memo 'Generic Resources'

> must directives be maintained? are these the same fixed resource?
>
>  http://www.rfc-editor.org/rfc/rfc2616.txt
>  ftp://ftp.rfc-editor.org/in-notes/rfc2616.txt
>
> this one too?
>
>  http://www.w3.org/Protocols/rfc2616/rfc2616.txt

You should be able to answer all these questions based on
the definitions. What are you asking? I guess I need to make
it clearer that the mereological relations from fixed resource
to content and directive-set are functional. Fixed resources are by
design the simplest possible 'information resources'.

We agreed earlier in this project not to worry about time, as it's
an unsolved problem for RDF with no special relationship to this
project.

>> Term: 'representation' (class)
>>  Def: Similar to 'fixed resource' but independent of provenance.
>>  That is, the identity of a 'representation' is determined by the
>>  content and directives.  In case two processes coincidentally result
>>  in the same content and media type, we would say there are two fixed
>>  resources but that the two fixed resources 'have' the same
>>  representation.
>>  Similar to: RFC 2616 'entity', AWWSW 'representation'
>>
>>  Note: JAR and Pay Hayes both oppose making a distinction betwen
Pat
>>  fixed resource and representation.  The distinction is included at
>>  TimBL's insistence.
>
> if fixed resource consists of only content + directive, and all the
> properties are identical, how would you ever know they were different in the
> first place? One must first be able to determine you have two distinct
> things - I may be missing something here.

You might not know they were different. That doesn't mean they're the
same. A statement can be true and unknowable at the same time.

But one way to tell would be to inspect the apparatuses that produce
them, and see whether their apparent identity is coincidence or not.

For example, if resource 1 had been connected to a thermometer, and
the content gives the temperature, and resource 2 had been connected
to a spring scale, and the content gives weight, then TimBL would say
they're different fixed resources, even if their 'representations' are
bit-for-bit identical. This difference is what we've been calling
"phlogiston" (I tried "intent" but was browbeaten).

>> Term: 'has fixed representation' (property)
>>  Domain: fixed resource
>>  Range: representation
>>  Functional.
>>  Def: The representation is a projection of the fixed resource,
>>  i.e. shares its content and directives.
>>
>> Term: 'is authorized for' (data property)
>>  Domain: 'representation'
>>  Range: URI (xsd:anyURI)
>>  Def: The agent that legitimately controls the URI (its "owner")
>>  authorizes the representation for inclusion in an HTTP 200 response
>>  to a GET request with the URI as target URI.
>>  Definition source: HTTPbis part 1 section 2.6.1
>>
>>  http://tools.ietf.org/html/draft-ietf-httpbis-p1-messaging-12#section-2.6.1
>
> if we're still in the domain of HTTPbis here...
> - assuming 203 Non-Authoritative Information comes in to play here?

Well, the content is by definition not authorized, so no. I'm only
talking about authorized content (200).

> - Content-Location which differs in domain or scheme?

Semantics of Content-location is future work (it's similar to 30x,
which on Tuesday we decided was a separate project)

>> Term: 'is served fixedly from' (data property)
>>  Domain: 'fixed resource'
>>  Range: URI
>>  Def: R is served fixedly from U iff R's representation is the only
>>  representation authorized for U.
>>  Similar to: log:URI ?
>
> define R?

Universally quantified. I'm speaking mathematician-ese.  More explicitly:

Def: a fixed resource R is served fixedly from a URI U iff R's
representation is the only
representation authorized for U.

or:

Def: For all R and U, R is served fixedly from U iff R's
representation is the only
representation authorized for U.

This definition could also be expressed in OWL-DL.

> I've been trying to find a way to prove that one and only one representation
> exists for U (using httpbis) and can't - does that affect anything? (I can
> provide a list of every case where I thought it may be possible to assert
> this, and the reasons why it can't be asserted)

Not sure what you're asking. The note following is explicit.

>>  Note: 'is served fixedly from' is not functional under this definition.
>>  If
>>  R is served fixedly from U then so is any fixed resource S that has
>>  the same representation.
>>
>>
>> Now we can stipulate that fixed resources, or at least some of them,
>> can act as subjects of metadata, by interpreting the content
>> (as modulated by the directives) appropriately.  In particular:
>>
>> - RDFS - can participate in rdfs:seeAlso, rdfs:isDefinedBy, etc.
>
> does that follow?

No, but it's (a) consistent, (b) useful, (c) widely deployed, and (d)
the purpose of the exercise.

Maybe we can come up with axioms that help formalize this informal
idea, but I'm not sure that would help.

>> - FOAF  - every fixed resource is a foaf:Document, so can participate
>>  in foaf:homepage, foaf:topic, etc.
>
> so foaf:Document is the class of all fixed resources?

I never said that.  I just said that 'fixed resource' is a subclass of
'foaf:Document'. (That can easily be expressed as a formal axiom.)
Obviously there are other foaf:Documents - those that change through
time, or have ads, or vary by user-agent or conneg or authentication.

>> - Dublin Core  - fixed resources can have dc:creators, etc.
>
> can have dc:hasPart & dc:isPartOf?

Why not?

> dc:modified? (fixed resource can be
> modified?)

No (for some fun, see
http://www.balisage.net/Proceedings/vol3/print/Renear01/BalisageVol3-Renear01.html
)

>> - CC REL - some fixed resources are legally protected by copyright
>>   and can be licensed (xhtml:license property)
>>
>> We can express a form the httpRange-14 rule:
>>
>>  When exactly one representation is authorized for U,
>>  (a) everyone is urged to use U to name only some particular fixed
>>  resource served fixedly at R (which one is not clear, but probably
>>  depends on the process by which it was created or chosen by U's
>>  owner),
>>  (b) absent information about which fixed resource is involved,
>>  just assume that U names some unknown fixed resource served fixedly
>>  at R.
>
> so conneg would be impossible, if I read correctly?

Yes, for fixed resources.  See Tim's note.

> generally, I'm reading this as, every fixed resource is 1-1 with a single
> "representation" from the time of creation till the end of time. is that
> correct?

Yes. See Tim's note.

> Best,
>
> Nathan

Will try to clarify things for the next draft. Thanks.

Jonathan

Received on Friday, 21 January 2011 21:39:57 UTC