Re: Aggregation: simple proposal

On 4 Jan 2013, at 20:29, John Arwe <johnarwe@us.ibm.com> wrote:

> Henry, I'm going to start drafting a proposal on issue 37 ("the model") to try and get things moving.  I was going to incorporate your simple proposal from [1] and build on it, but I find myself asking many clarifying questions about its intent.  Andy seems to divine your purpose, but I have a nice linear reading style that dumps on any ambiguity. 
> 
> > >> 1. Create a aggregation container
> > >> ----------------------------------
> > >> 
> > >> POST your collection to the container
> > >> 
> > >> <> primaryTopic <#wk> .
> > >> 
> > >> <#cnt> a WeakContainer;
> > >>     :contains <http://remote.org/resource>,
> > >>         <localRes> .
> > >>     
> > >>     This creates a resource named say <resourceC> which defined <resourceC#cnt> .
> 
> Q1.1: What is the purpose of the primaryTopic triple?  It seems like a distraction from the main purpose.  I wonder if you are mixing "before-state of the LDPC" with the POST request entity body's content.  When I write this out pedantically I intend to show the LDPC before and after each interaction, hopefully to make it unambiguous. 

The point is to distinguish a collection that in the LDPC case is something like a directory, and so a real web resource if you want,
whereas we are creting a collection that is like another object - like giving a URL to a cat perhaps. An LDPC is something you
POST to create an LDPR. The aggregation collection I am talking about is not something you can post to.

> 
> Q1.2: You use both "collection" and "container" ... should I take "collection" to mean 'aggregation' and interpret container 'literally'?  [later add] You also use "aggregation container" in the title; I assume that's the same as "aggregation" elsewhere in this snip. 

I wrote it quickly.
LDPC as they are now are things you can POST to create new resources. The collections I am thinking of here are not such things.



> 
> Q1.3: Am I correct in thinking that the WeakContainer type is merely a mnemonic convention to convey the 'aggregation' intent? 

yes

> 
> Q1.4: You are applying the rule "null relative URI denotes the URI assigned to the new entry by the container's create implementation"?   So far I've only seen that discussed for <> as a "special case", although I certainly anticipated that once formally drafted it would cover <#cnt> in a similar way, but want to be sure that's your intent as well.  Using that interpretation, I'd say that <resourceC> is the server-assigned -portion- of the URI (the absolute URI chosen by the server to allow transformation of <#cnt> into the absolute URI <resourceC#cnt>). 

yes.

<> refers to the document created when you POST to a LDPC.
<#me> and stuff like that refer to identities defined in the document <> .

> 
> Q1.5: The "This creates..." sentence ties me in knots.  What would you expect in response to GET resourceC , assuming the server added+changed no triples from the POST representation?  Based on my theory about the answer to Q4, I'm guessing you expect back: 
> <> primaryTopic <#cnt> . //<-had an error here
> <#cnt> a WeakContainer;
>        :contains <http://remote.org/resource>, <localRes> .
> 
> I agree that would all be "legal", btw.  Just being sure I have your intent. 

yes, so of course if the server is <http://example.com>
and the container is <http://example.com/cnt/>
which created  <http://exmple.com/cnt/weakColl>
then the above is in full RDF

<http://exmple.com/cnt/weakColl> primaryTopic <http://exmple.com/cnt/weakColl#cnt> . 
<http://exmple.com/cnt/weakColl#cnt> a WeakContainer;
       :contains <http://remote.org/resource>, 
                      <http://exmple.com/cnt/localRes> .

> 
> Q1.6: For a "simple" proposal, why start with both a collection and a container (regardless of the answer to Q1, "clearly" you intended to distinguish by using distinct words)?  Versus, for example, saying "this is how an LDPC could be used, without spec changes, to accomplish aggregation"?  

yes, I intended to distinguish both. There is good reason to argue that LDP Containers (LDPC) are very different from weak collections. Atom Protocol I think has a similar distinction between atom:Feed and app:Collection

 https://tools.ietf.org/html/rfc5023#section-8.3.3

The Collection is something that is defined by the protocol, as something you can POST to to create new server resoures, and GET, and DELETE to delete all of its contents. That is not true of say a normal atom:Feed, even thought they are very close.  So similarly here we can distinguish one from the other.

> Are you asserting that the use of "indirection" in this way is required to "implement" aggregation using LDPC rules as they currently exist?  [again, just getting your intent - I'm not asserting dis/aggreement at this point] 

No I think my idea was to keep those separate. I just then chose WeakContainer, because I did not know what to call it, and I did not really think that this was a further debate one could have. The point at present was just to show how one could have a simple and limited concept of a strong container, and still have all the freedom needed to do aggregation.

> 
> Q1.7: For a "simple" proposal, why start with a case where a single HTTP resource manipulates multiple (2) RDF resources (subject URIs)? 
> 
> 
> > >>     2. Delete your container
> > >>     send an HTTP delete message
> > >>       DELETE resourceC HTTP/1.1
> > >>     or change the container as below by removing the <#cnt> from <resourceC> 
> 
> Q2.1: You say delete the *container*, but you use resourceC as the request URI.  In your 1., 'container' was the pre-existing resource you posted to in order to *create* <resourceC>.  Did you mean to say delete the aggregation/collection instead?   Same occurs in 3.

given my above URLs this should be 

DELETE http://exmple.com/cnt/weakColl HTTP/1.1


> 
> 
> > >>     or change the container as below by removing the <#cnt> from <resourceC> 
> 
> Q2.2: I think by this you mean either using the method of (3) with a request entity body consisting of ONLY the primaryTopic triple, and/or using the method of (4) with DELETE { <#cnt> :contains ?o }  [my ?o syntax might need some adjustment, working from memory] 

yes. But the primary topic you are right here is not essential ( or at least it may be, but it would need to be argued )

yes, my suggestions was just you could remove all triples mentioning the collection. That would be equivalent to deleting it.

Hope that helps :-)

	Henry

> 
> 
> 
> 
> [1] http://lists.w3.org/Archives/Public/public-ldp-wg/2012Dec/0112.html
> 
> Best Regards, John
> 
> Voice US 845-435-9470  BluePages 
> Tivoli OSLC Lead - Show me the Scenario 
> 
> 

A short message from my sponsors: Vive la France!
Social Web Architect
http://bblfish.net/