[Recent Entries][Archive][Friends][User Info]
| Time | Text |
|---|---|
07:01 pm![[info]](http://lj.rossia.org/img/userinfo.gif){kind=small} apophates[Link] | http://apophates.livejournal.com/11 On the scale of the World Wide Web, any form of centralization would create immediate bottlenecks, in terms of network throughput and server capacity. This need for a flat, non-server-centered architecture will be even stronger on the Semantic Web. Of course, the same physical load-balancing arguments hold as on the current Web, but the Semantic Web adds a new argument in favor of a P2P-style argument. On the Semantic Web, any server-centered architecture will not only create physical bottlenecks, but as communication relies on the use of ontologies will also create semantic bottlenecks. Since the semantics of information will be explicit (or at least more explicit) on the Semantic Web, any single server will in a way “impose” a particular semantic view on all its clients. This will have undesirable consequences, both in terms of the pluriformity of the available information, as well as in terms of the size of the central ontology that such information-servers would have to maintain. Instead, a P2P-style architecture will be able to avoid both the physical and the semantic bottlenecks. Different semantic views, expressed in terms of different ontologies, will be provided by many peers in a flat network of peers, each employing their own local, small ontology. Of course, this increased flexibility comes at a price: such “different semantic views, in terms of different ontologies” create a significant data-integration problem: how will these peers be able to communicate if they do not share the same view on their data? In the remainder of this paper, we propose an approach where the communication between peers relies on a limited shared vocabulary between them. This replaces the role of the single virtual database schema that is the traditional basis for solving information exchange problems... |
| Reply: | |