- From: Alexandre Bertails <bertails@w3.org>
- Date: Fri, 18 Oct 2013 11:46:01 -0400
- To: Andy Seaborne <andy@apache.org>, public-ldp-patch@w3.org
On 10/18/2013 11:05 AM, Andy Seaborne wrote: > On 18/10/13 15:24, Alexandre Bertails wrote: >> On 10/18/2013 10:13 AM, Andy Seaborne wrote: >>> On 18/10/13 03:57, Alexandre Bertails wrote: >>> >>>> General remark: Linked Data (in LDP) is different from general RDF: >>>> the data lives in "small" HTTP documents, not in "big" RDF store. >>> >>> Hmm - collections have the potential to be large and, in general, >>> planning on "small" seems to fail the test of real use! >> >> Collections as in LDPC, yes, that is true. I was talking about LDPRs. > > If LDPC are a subclass of LDPR ... :-) It's true that the ontology says so, but I don't think that there is a subtyping relationship in practice. The spec itself does not define an LDPC as a refinement for an LDPR: it discriminates the two cases. Would one implement an LDPC class y extending LDPR? Would one expect all the tests for an LDPR to hold for an LDPC? You bring an interesting question though: can I PATCH an LDPC? And if I can, do the LDPR data live in named graphs or in the default graph? The spec doesn't say anything about that (would be good though). > >>> If it were truly small, than PATCH is not really needed - use PUT and >>> etags. >> >> I'd be happy to add this approach in [2]. Is there some text somewhere >> about how exactly one can use ETags for LDP PATCH? > > It is just normal use of etags and PUT - plain web stuff. Ok, I thought you were using ETags in relation with the skolemization. Like, a bnode label would be tied to a specific ETag or something. Also, even "small", the PATCH would still be very important. [snip] >>> > * blank nodes are system dependant, so not well specified in the case >>> of LDP >>> >>> Skolemization? >> >> Sure. But the specification in its current form says [[ Blank nodes >> are recorded as _:id where id can be the system-internal identifier >> for the blank node ]], but says nothing about skolemization, does it? >> >>> >>> As the RDF patch docs discusses, both doc-wide scope (RDF style) and >>> store-scope are needed for different usages. >> >> I understand. So here is my question: what scope does RDF PATCH >> address? > > (A.4) For others, Andy meant http://afs.github.io/rdf-patch/#alternative-bnode-syntax [[ Both RDF-style (document scoped labels, new blank nodes created) and store-scoped (existing blank nodes) are needed. This may require two syntaxes to distinguish. Should store-scoped blank nodes have a special encoding like <_:label> and the _:label form behave like RDF syntax? ]] > > From other feedback, it looks like it needs both. > > store-scope to do changes, as all the proposal have to address some how. > There are two appraoches - find the bnode, and name the bnode. What is store-scope in the case of LDP? There is no central service endpoint so I don't understand where this would be useful. For PATCHing through an LDPC? > > RDF-patch happens to mention <_:label>, not http skolemization but that > is an artifact of history and a laziness in writing > <http://host/skolemId/ABCDEF> all the time. I still don't like skolemization. Basically, JSON-LD would look awfull if we were to skolemize bnodes. I think that bnodes and their current support in Turtle and JSON-LD are a feature, not something we should try to avoid. > > doc-scope, like RDF syntax, for new data. > > One caution: finding bnodes avoids the need to name then in the access > operations ... but finding them is really quite tricky in practice > unless the data is well behaved (= planned for this sort of use, e.g IFP). That's something I'd like the group to discuss. Many people argued that in the case of LD, this is not an issue. That IFP, nailed nodes and URI+property path would allow us to bind most interesting bnodes. > > A test case is RDF list manipulation as is deleting molecule-like > structures uses for structured values. Pierre-Antoine's proposal does handle rdf:list. What do you think of his approach? Also, I'm not sure to understand what "molecule-like structures" are. Alexandre. > > Andy > >> >> Alexandre. >> >>> >>> Andy >>> >>> >>> >>> >> >> > > >
Received on Friday, 18 October 2013 15:46:02 UTC