Re: Considering a PATCH model for LDP

Because you are labelling/naming the graph in a local context (the file) 
with a global name.

I disagree with:
A pre-defined named graph that represents the nodes to match or to 
insert is sufficient.

Your 'oslc:deleteGraph' graph isn't my 'oslc:deleteGraph'.

What description would I get if deference GET oslc:deleteGraph?

You are proposing using fixed, well known names as syntax to construct 
the change operation. Using keywords seems more appropriate.

- - - - - - - - - - - -

All changeset proposals need also to deal with deleting bNodes -

oslc:deleteGraph {
_:b1 dcterms:title "A bad link title";
    rdf:object <http://myserver/mycmapp/bugs/1235>;
    rdf:predicate oslc_cm:relatedChangeRequest;
    rdf:subject <>;
    a rdf:Statement.

in TriG does not delete a triple with property dcterms:title.  _:b1 is 
label just within the file, and not the resource. On parsing you get 
different bNodes each time the file is read.

What's the proposal here?

DELETE DATA does not allow bNodes in the data specifically because of 
the meaninglessness.

You may which to consider DELETE WHERE {} which allows named variables 
(but not bNodes).  DELETE WHERE {} is a shorthand for

DELETE {template} WHERE {pattern}

with template=pattern.


On 11/03/13 12:48, Steve Speicher wrote:
> On Sun, Mar 10, 2013 at 3:55 PM, Andy Seaborne
> <> wrote:
>> The use of global names for the actions on the graphs can be avoided by
>> including a manifest in the default graph (the way the global names are uses
>> is really predicate-like).
>> { <> ldp:insert :g1 ;
>>       ldp:delete :g2 }
>> :g1 { ... }
>> :g2 { ... }
>>          Andy
> Thanks for the feedback Andy.  You seem to imply that "global names
> for the actions" is a bad thing, can you elaborate why this is a
> problem and what you propose improves it?  I think I know the answer
> but I'd rather not assume.
> Thanks,
> Steve
>> On 09/03/13 23:34, Ashok Malhotra wrote:
>>> Hi Steve:
>>> We have PATCH for JSON
>>> and PATCH for XML:
>>> Wouldn't your proposal be better characterized as PATCH for RDF?
>>> So, we should have the RDF folks look at it.
>>> All the best, Ashok
>>> On 3/9/2013 9:35 AM, Steve Speicher wrote:
>>>> We have a number of issues related to PATCH'ing resources:
>>>> ISSUE-12 (closed) Can HTTP PATCH be used for resource creation?
>>>> ISSUE-17 changesets as a recommended PATCH format
>>>> ISSUE-27 Should the PATCH method be used, as opposed to POST with a
>>>> given mime type?
>>>> I have drafted something very simple that meets most of OSLC's simple
>>>> use cases that I would like to use as a basis for discussion on a
>>>> model for PATCH [1]. It separates the model from the document (format)
>>>> used mostly.  It takes an approach that doesn't require SPARQL Update
>>>> but shows how it can be used.  The patch document can be any quad
>>>> format.
>>>> Feedback welcome on this independent of usage within LDP as well.  I
>>>> realize the proposal is incomplete and apologize for that -- I thought
>>>> there was value in sharing what I have so far.  I'm currently working
>>>> on some additional validation of this approach as well.
>>>> [1] -
>>>> --
>>>> - Steve Speicher

Received on Monday, 11 March 2013 13:25:48 UTC