6 Conclusion: integration of location-management with service discovery infrastructures

Service discovery infrastructures such as SLP[15], Jini [16], UPnP [17], Salutation[18] and UDDI[19] attempt to move up from purely network-based addressing, to account for higher-level descriptions of networkable entities. They provide a bootstrap mechanism that makes possible the dynamic, spontaneous hookup of services and devices in ubicomp environments. They use, in a centralised or distributed fashion, a generalised lookup service that may build upon and subsume the more specialised naming, trading or directory services provided by underlying middlewares and protocols. As such, both direct, inverse and combined location queries should be directly supported by these infrastructures.

We are still far from this objective, if only for the lack of widely adopted representation standards for location information and associated data, at all the different levels described before. Programmatic interfaces for services and their attributes, such as used by Jini , are both highly expressive and close to implementation (that is why they were chosen in the above-described implementation), but place a probably too high bar on interoperability. Declarative XML-based interfaces are a more pragmatic solution for minimal interoperability and could make it possible to interoperate at all levels of the architecture described above, provided standard DTDs, (preferably schemas), become widely adopted for each of these. There is already a substantial amount of work on this, which unfortunately has led to the definition of different and incompatible DTDs (or, worse , binary formats) by all the different consortia and standardisation bodies interested in this topic, from geographers to mobile telecom operators, to specialists of 3D graphics[2]. For the lower layers where relevant ontologies are fairly straightforward and could be agreed on by all parties, this multiplication of overlapping and competing would-be standards could have been avoided. As usual, a shakeout will let one of these emerge, but this may take some time. For the structural and semantic layers, the nature of the concepts to be manipulated is much more open and an a priori definition more difficult. Here, as in the general issue of semantic service discovery [20], a meta-level specification of location ontologies should be possible: we could have, for example, different models for intra-building location based on different kinds of architectural entities, or culturally-dependent street/precinct hierarchy models for describing location in an urban area.