AWWSW -- 01 Apr 2008

Relationship between a thing and an HTTP response

<jar> http://lists.w3.org/Archives/Public/public-awwsw/2008Mar/0015.html

jar: It's beginning to make sense to me.

<jar> http://esw.w3.org/topic/AwwswThingResponseRelationship

<Stuart> Re: By RFC2616, 200 means that "an entity corresponding to the requested

<Stuart> resource is sent in the response"

<Stuart> I may choose to disagree with the way that is expressed.

Stuart: My homework conveys the request and the response.

http://lists.w3.org/Archives/Public/public-awwsw/2008Mar/0015.html

Stuart: I'm inclined to think one is asking a question of "the Web" and "the Web" responds.
... So I misread jar's comment.
... I posted a transcript and wrote a little RDF to model the operation and response. Also made a rule with an antecedent to pick out the date, content, etc., and generates conclusions, including that the response conveys the state of the resource.
... What the bits encode is what you want to get at. It's comparing two resource states as being the same.
... And I mentioned a few things that are not concluded, such as that the document is invariant, it is the US constitution, etc.

Sounds like we might want to use line numbers when we submit code. :)

alan: Does that URI denote a class or instance?

Stuart: An instance.

Alan: The page http://lists.w3.org/Archives/Public/public-awwsw/2008Mar/0015.html also says "The troop surge in Iraq is over a year underway". Is that part of it?

Alan seems to be asking what exactly the URI denotes.

Alan: The document changes over time, e.g., it has a counter.
... But I could make statements about it as containing the constitution.

dbooth: In an HTTP, you need an actual URI, not just <http://www.usconstitution.net/const.html>.

Stuart: If you put angle brackets, I mean the resource associated with that URI.
... If it had been the lexical string I would have put quotes around it.

alan: Needs to be the URI string, not angle brackets.

Jar: That same resource could have multiple URIs, and you might get different responses from different URIs.
... You've said that the resource as some particular state -- a time-sensitive conclusion. It may be true for some period of time, then may no longer be true. That's different from my approach which says there was some time when some particular state was in effect, which will always be true.

dbooth: but hasState doesn't preclude it from also having other states (with different times)
... The word "state" or "hasState" seems misleading, because it what you get from the resource may not be the state of the resource.

alan: This thing has states over time and you get one of these states back. That's not the same as a MS Word doc that doesn't change over time.

<jar> stuart: We don't know much from the http interaction itself - just that it's a Document

stuart: point of second rule, if I picked on a W3C TR I would have more than one URI through which i could retrieve a rep of the most recent version. ONe has a date embedded and the other doesn't. They're different resources, but at some times they communicate the same state.

<jar> dbooth: 'state' is misleading

<jar> dbooth: it's not the 'state' of the resource, it's just some data that was sent back

alan: there's no constraint on what a valid rep could be.

dbooth: that's a feature, not a bug!

stuart: what is it a representation of? could be of the resource or of the state of the resource.

<jar> stuart: a representation is a representation of what? the resource or the state?

alan: i could ask you what time is it, and you could answer "potatoes".

dbooth: yes, so use a different resource, that is more helpful in telling the time.

stuart: Roy Fielding says that what makes something "resource"ful is its consistency over time. If it isn't consistent, it isn't useful, few will use it, and it will die.

alan: So that means we don't say what resources are.

jar: The site owner gets to say how the URI should be used as a name. Since the HTTP protocol isn't saying how the name should be used, you need to look elsewhere.

dbooth: I partly disagree.

jar: the only thing you can conclude is that there's some undefined relationship between the resource and the response.

<jar> 10 minutes left

dbooth: stuart, why do you provide a rule for comparing states?

stuart: this is to be able to indicate that two different resources may have the same state even if they are different resources.

<jar> alanr: if it doesn't change it doesn't have state

<jar> stuart: sure it does - the state may not change, but that's ok

<jar> shall we continue past 10:00?

<alanr> k

<jar> alanr: David, can you unpack a bit your dislike of the word 'state'?

<jar> dbooth: The information coming back is completely arbitrary - unconstrained, the server gets to decide

<jar> alanr: Then there is no relationship

<jar> dbooth: Yes, there is, the data was sent back on behalf of that resource

<Stuart> dbooth... just about... but it also to note that in general a resource is not it's current state, that state can change AND that juts because 2 resources have the same state (possibly at the same or at different times) does NOT allow you to conclude that they are the SAME resource.

<jar> dbooth: Well, it's a very weak relationship, but a relationship nonetheless

<jar> alanr: A relationship between the *name* and something is not one between the thing and the something (e.g. me and my putative hotel room)

<jar> dbooth: http doesn't tell you very much about the relationship between the response and the resource

http://lists.w3.org/Archives/Public/public-awwsw/2008Mar/0016.html

<jar> dbooth: What you *can* determine is in my response to the homework assignment

<jar> alanr: Stuart, if someone looked in a state, and didn't find the constitution, then a mistake would have been made, right?

<jar> stuart: Someone has decided to deploy the resource, and if they decided that is should be the constitution, then yes, a mistake would have been made

<jar> stuart: i haven't published that intent. 3rd parties may observe that it appears to be the us constitution, and revisit it, but maybe it will disappear

<jar> alanr: If a statement of the type was published, then it would be possible for *others* to say a mistake had been made, yes?

alan: We could have an extra method that says what kind of thing the URI denotes, and it could say that it's supposed to be the constitution, and then if the response doesn't contain it, then we could say that it is erroneous.

stuart: You're not going to find out that intent from http. You might with a Link header that points to a description.

<jar> it's 10:11. wind up?

I agree with Stuart.

<jar> alanr: Maybe this is too complex a case?

dbooth: in some sense the IR case is more difficult than the non-IR case, because with a non-IR you have what I've been calling a URI declaration that explicitly says what the URI denotes. We don't (yet) have such rules for IRs (though my homework proposes one).

<jar> alanr: Looks like you don't learn much of anything from a 200 response.

<Stuart> The response to a GET is *NEVER* the reference resource.

<jar> stuart: Yes you do - you can read the page and it tells you things -

<jar> jar: But you only know about the responses, not the resource

<Stuart> no... you get a representation of the current state of the resource.

<jar> Sure, but is there anything I can say about the resource itself? Not without an independent statement.

<Stuart> The resource has future state which may or may not be different... you do not get all that future potential.

<jar> dbooth: Encourage people to look more closely at writings about URI declarations

- DRAFT -

AWWSW

01 Apr 2008

Attendees

Contents

Relationship between a thing and an HTTP response

Summary of Action Items

Scribe.perl diagnostic output