Re: composition and aggregation

On 29 Jan 2013, at 18:23, "Wilde, Erik" <Erik.Wilde@emc.com> wrote:

> hello henry.
> 
> On 2013-01-29 17:28 , "Henry Story" <henry.story@bblfish.net> wrote:
>> When I do an HTTP GET on your Container I need to be able to distinguish
>> which things are contained in the container and which things are linked
>> from it. So assuming your http://my.example/wilde/ container returns
>> G1 = {
>>     <> a ldp:Container;
>>        rdfs:member <friends>, </joe> .
>> }
>> Then how would an agent know which of <member> or </joe> was
>> a contained element, and which one was linked?
> 
> it wouldn't, but why would that be a problem? notice that in the proposal
> i have made, <friends> probably wouldn't be in there, because members
> always would be managed by the server. so you would get something like
> this:
> 
> <> a ldp:Container;
>   rdfs:member </jim>, </joe> .
> 
> 
> and then when GETting /jim, the content about jim would be embedded in
> jim, whereas when GETting /joe, the content would be linked, and you would
> need to GET /joe's content with an additional GET. we could of course, and
> we probably should, expose the content triples in the collection listing
> already, so that we would reduce the number of required GETs.
> 
>> Now a client can tell which of those resources if deleted from
>> <http://my.example/wilde/> would in fact be removed. The answer
>> is <friends> would be removed. Also you know that if you send
>> a DELETE to <friends> it will be removed from the ldp:Container.
> 
> in my proposal, both /jim and /joe get deleted, so there's no need to make
> any statements about who gets deleted. everybody gets deleted. the only
> thing that will not get deleted is /joe's content, but it will lose its
> status as being /joe's content anyway and simply will continue to happily
> live as a web resource.
> 
>> So perhaps you POST
>> P1 = {
>>  <> ldp:content <http://bblfish.net/> .
>> }
> 
> that, plus all the other metadata that's required by the LDP protocol,
> plus all the optional metadata i am planning to store.
> 
>> to  <http://my.example/wilde/> then a new GET on
>> that resource will return
>> G3 = { 
>>   <> a ldp:Container;
>>       ldp:contains <friends>;
>>       ldp:link </joe>, <http://my.example/wilde/> .
>> }
> 
> now i am very confused. http://my.example/wilde/ is the container, right,
> because you POSTed an entry to it. upon success, you will get a new
> resource, maybe http://my.example/wilde/42, that will contain the graph
> (maybe with whatever our protocol says we want to add, such as a link to
> the container) you POSTed.
> 
> have you made an error here when adding ldp:link
> <http://my.example/wilde/>? i really don't understand why you would GET
> this when GETting the container.
> 
>> Ok so for an embedding example now:
>> Perhaps you post (forgetting namespaces for the moment )
>> 
>> P2 = {
>> <> ldp:content [ a animal:Cat;
>>                foaf:name "kittykitty"
>>                foaf:depiction <img/kity.jpeg> ]  .
>> }
>> So now because there is additional information on the blank node that
>> represents an animal,
>> Your idea is that a new GET on <http://my.example/wilde/> will return
>> G3 = {
>>   <> a ldp:Container;
>>       ldp:contains <friends>, <kitty>;
>>       ldp:link </joe>, <http://my.example/wilde/> .
>> }
> 
> you're assuming the server assigned "kitty" as the entry URI? then yes for
> that part, but again i don't understand the container resource linking to
> itself.
> 
>> Where a GET on <http://my.example/wilde/kitty> will return perhaps
>> K1 = {
>> [] foaf:name "kittykitty";
>>    foaf:depiction <img/kitty.jpeg> .
>> }
> 
> nope, we GET an entry, so we GET the entry's metadata (since you created
> none, maybe just whatever the server assigned, let's say a creation date),
> plus the content still represented in the same way as you POSTed it. K1 is
> an invalid LDP entry, because each LDP entry MUST contain exactly one <>
> ldp:content triple.
> 
>> Anyway, is this what you are getting at or am I misunderstanding what you
>> are thinking of? 
> 
> i cannot really tell because i am not fully following the examples you
> have included. but it seems you have a fairly different model in mind from
> the one i am trying to explain.

Well perhaps you need to explain what you want to do then, by writing this out 
in detail in Turtle. Because I don't understand what you want either. I can
guess, but clearly that's a waste of time.

> 
> cheers,
> 
> dret.
> 

Social Web Architect
http://bblfish.net/