- From: Wilde, Erik <Erik.Wilde@emc.com>
- Date: Wed, 31 Oct 2012 18:37:02 -0400
- To: Olivier Berger <olivier.berger@it-sudparis.eu>, Arnaud Le Hors <lehors@us.ibm.com>
- CC: "public-ldp-wg@w3.org" <public-ldp-wg@w3.org>
hello olivier. On 2012-10-31 04:03 , "Olivier Berger" <olivier.berger@it-sudparis.eu> wrote: >In the case where containers get "external" resources added to them >(aggregation, weak links, etc.), I think it would be quite interesting >to consider that eventually these resources could get notified that they >just got added to a LDP C. >Similarly, when a resource gets deleted, the container could be notified >about that to eventually remove it from its members. >Does it make sense wrt the LDP scope ? i would hesitate to do that. like i tried to point out in my original post, links often are more "service" that they are "data", so trying to persist them as "data" may not be such a great idea (and i think this is what arnaud initially meant when i talked about the headaches of managing bidirectional links). so instead of treating links as data, in loosely coupled architectures you'd rather design services that other can ask at runtime about reverse links. for example, it would be a quite a headache for a big blog to be notified every time somebody mentioned a blog post, and then having to store all these backlinks, and maybe even testing whether they still are alive. much better to have a service (such as google) that you can go to, whenever it matters, that aggregates and manages the links, and at any point in time will tell you who is linking to your blog post. trying to persist all relationships between resources on the web, in particular those that are not under your authority, is a scalability and maintenance nightmare, and might be something you maybe shouldn't even try. knowledge on the web always is partial, and that's a fact of life. iff we wanted to tackle this, i'd much rather have the kind of "linking service" i outlined above, rather than putting the burden of managing incoming links on everybody's data directly. i guess RDF would be a very good foundation for that, you would just have a service you'd ask about triples where some of your resources are the link target, and then you'd get a bunch of triples you could mix in with yours, but it probably would be a very good idea to not persist those, but to treat them as a snapshot that could change at any time. michael hausenblas might want to pitch in here, because maybe there could be some synergy between this idea and VoiD, which as i understand it is a more static model of how "resource graphs" could be interlinked, but maybe could be extended to be something more dynamic (and i am sorry, michael, if that is an absurd misrepresentations of what VoiD is all about). cheers, dret.
Received on Wednesday, 31 October 2012 22:38:50 UTC