Re: MKCOL for making collections

On 22 Jan 2013, at 16:55, Alexandre Bertails <bertails@w3.org> wrote:

> On 01/22/2013 10:53 AM, Henry Story wrote:
>> 
>> On 22 Jan 2013, at 16:35, Alexandre Bertails <bertails@w3.org> wrote:
>> 
>>> On 01/22/2013 10:17 AM, Steve Speicher wrote:
>>>> On Tue, Jan 22, 2013 at 10:03 AM, Henry Story <henry.story@bblfish.net> wrote:
>>>>> ISSUE-36: Can applications create new containers?
>>>>> 
>>>>> We cannot make a collection by POSTing a doument on a collection, since that
>>>>> creates a resource. We therefore would need a different HTTP Method to do this.
>>>>> 
>>>> 
>>>> Why can't we do this?  A Container/Collection IS a resource.  So
>>>> therefore POST'ing the representation of it seems like the most
>>>> obvious way to create one.
>>> 
>>> I agree with Henry's premises: this way of creating containers should
>>> be discussed. The reason is that this strategy forces you to look into
>>> each RDF document that is POSTed, to search for the triples saying
>>> "hey, by the way, I'm an LDPC!".
>>> 
>>> It's not that it's impossible, it's just very annoying in practice. So
>>> it's at least not perfect.
>> 
>> Thanks :-) I was just writing this up.
>> 
>> Next interesting enough would be the question as to whether a link would
>> do to a special resource for creating a collection in your collection.
>> You could have
>> 
>> <> ldp:createCollection <xxx> .
> 
> You wouldn't do that anyway. Something like that is more appropriate:
> 
> <> a ldp:container .

I was considering the case where someone did not want to use MKCOL
but wanted a relation to a "form" where he could create a collection in 
this collection. That's what the argument below was trying to make
explicit.

Your suggestion of just having

<> a ldp:Container .

all by itself is of course the right thing to do if you accept MKCOL.

> 
> Alexandre.
> 
>> 
>> But now what is xxx ? It cannot be something you can add attributes to
>> like <xxx?name=pix> to create a collection and that then you can send
>> a GET to. Because GET is indempotent and you'd be in danger of crawlers
>> creating collections all over the place.
>> 
>> If is something you POST to to create a collection, then is it not
>> an ldp:Container? But if it is an ldp:Container then it's a bit odd
>> that you are POSTing to another ldp:Container, in order to create a
>> new container in your collection.
>> 
>> My guess is that it is reasoning of that type that led to the creation
>> of the MKCOL method.
>> 
>> 
>>> 
>>> Alexandre.
>>> 
>>>> 
>>>>> I suggest MKCOL from WebDAV, since it is already defined.
>>>>> 
>>>>> 
>>>>>    http://restpatterns.org/HTTP_Methods/MKCOL
>>>>> 
>>>>> 
>>>>> The content of the body, could be Triples that describe things
>>>>> that can be put into the collection. As it happens that is not
>>>>> defined yet it seems, so we can define it.
>>>>> 
>>>>>   Any thoughts on that?  I'll try implementing that.
>>>>> 
>>>> 
>>>> Seems like we'd have to dig into more of what the semantics of that
>>>> verb says [1] but feels a little like we'd have to adapt more of
>>>> WebDAV instead of just reusing the verb and giving it our own special
>>>> meaning.
>>>> 
>>>> [1] - http://www.webdav.org/specs/rfc2518.html#METHOD_MKCOL
>>>> 
>>>>> Henry
>>>>> 
>>>>> Social Web Architect
>>>>> http://bblfish.net/
>>>>> 
>>>> 
>>>> 
>>>> 
>>>> --
>>>> - Steve
>>>> 
>>>> 
>>> 
>>> 
>> 
>> Social Web Architect
>> http://bblfish.net/
>> 
> 

Social Web Architect
http://bblfish.net/