- From: Wilde, Erik <Erik.Wilde@emc.com>
- Date: Wed, 3 Oct 2012 15:16:59 -0400
- To: Arnaud Le Hors <lehors@us.ibm.com>, "public-ldp-wg@w3.org" <public-ldp-wg@w3.org>
hello arnaud. On 2012-10-03 11:53 , "Arnaud Le Hors" <lehors@us.ibm.com> wrote: >IBM has been struggling with the concept of two way relationships such as >describes+describedby. While it's convenient to have two links so you can >navigate from one resource to the other in either direction, maintaining >those > links adds a significant amount of pain and constraints. absolutely agreed. but there's nothing that tells you that you should actually store and manage those links. after all, all that we're talking about is how we can exchange representations that expose those relationships, and in many cases, these should be computed. but then again, if you talk about decentralized architectures, then the describedby links and described links will be manage by different authorities anyway, and in that case, there's no management problem (other than potentially broken links, but that's just a fact of life on the web). >This is clearly not the way the web of documents works. Anyone can add >links from their web page to other pages without expecting or requiring >reciprocity. This makes the web much more agile and resilient. But when >it comes > to data people tend to expect reciprocity and sometimes find >unidirectional links unacceptable. well, i certainly make links from my web pages to other pages unidirectionally, but then there's an amazing bunch of machinery out there crawling my page and inferring the backwards link, and then monetizing it. so yes, bidirectionality is not required, but in many scenarios, you have use cases for using established links in both directions, even though these may be served and consumed by different kinds of agents. cheers, dret.
Received on Wednesday, 3 October 2012 19:17:48 UTC