- From: Wilde, Erik <Erik.Wilde@emc.com>
- Date: Mon, 4 Nov 2013 12:11:27 -0500
- To: "ashok.malhotra@oracle.com" <ashok.malhotra@oracle.com>, "public-ldp-wg@w3.org" <public-ldp-wg@w3.org>
hello all. john already answered this, and it all comes down to that in REST, the resource model decided everything. if you model resources in a way that require you to lock some and interact with all of them and then release the lock, then most likely you have chosen a resource granularity that is not a good fit for the interaction patterns you want to support. i wrote about this a little while ago: http://dret.typepad.com/dretblog/2009/05/rest-and-rdf-granularity.html the main point is that even when you have an RDF model of something, you still haven't answered the hypermedia questions about the services on top of that: what are the interaction patterns you support, and on which "RDF documents"? that's critical for REST. On 2013-11-03, 16:58 , "Ashok Malhotra" <ashok.malhotra@oracle.com> wrote: > But I want to touch on a broader issue. With LDP we allow multiple > users access to a web resource, > and some of these users may be able to update the resource or some > part of it. This takes us into > database territory but we have steadfastly chosen to ignore > functionality like > - access control > - transactions > - locking > We need to consider these because it will come back and bite us in > the butt when, for example, users' updates are partially reflected >(transactions). john's claim and mine as well is that access control is out of scope (and i think we said so explicitly earlier on), while transactions and locking should be relegated to an appropriate resource granularity model. if an LDP server has problems with transactions/locking, then it probably has been set up in a way which exposes a resource model that's not a good choice. instead of changing LDP, the server then should change its resource model, so that interactions are actually stateless. cheers, dret.
Received on Monday, 4 November 2013 17:12:20 UTC