The pervasive computing vision was established around 1989 by work at Xerox on 'ubiquitous computing' and brought to public attention through Weiser's Scientific American article [6] and articles in Communications of the ACM [5,7]. The last few years have seen the emergence of suitable infrastructure technologies. Next generation Internet Protocol (IPv6) supports many aspects of pervasive computing, including the address space for large numbers of devices, and there is ongoing activity within the IETF (Internet Engineering Task Force) in support of mobility, ad hoc networking and automatic configuration. There has also been a huge increase in the availability, diversity and richness of digital content, which compounds the difficulties of content delivery and, more importantly, navigation in the increasingly pervasive setting. WAP (Wireless Application Protocol) addresses delivery of content, and techniques for dealing with adaptive and scalable delivery of multimedia content are attracting much attention (e.g. [4]).
This paper presents a prototype system that has been developed to explore hypermedia navigation in an ad hoc multimedia information space. The system assists users by automatically providing enriched linking between documents and other information resources brought together on-the-fly, for example in a meeting. We show a mechanism by which the Distributed Link Service (DLS) [1] could facilitate dynamic ad hoc linkbase creation, and propose that such an approach is useful in creating the pervasive information fabric; i.e. the middleware to support information applications in the pervasive computing context. Through the initial implementation, we highlight points for further research.
Imagine you are walking along a corridor, you look at a poster on the wall and the text and images on the poster prompt you to think of a number of associated things. Think of this as having a few collections of hypermedia links (linkbases) active in your mind. In our scenario, a wearable computer behaves in a similar way, with various personal linkbases, a linkbase associated with your current task and one associated with the building you are in. This highly dynamic, highly context-sensitive and ad hoc situation is what we are trying to realise through the hypermedia link services in our pervasive information infrastructure.
We extend this vision to a collaborative setting: you walk into a meeting room and sit down with several other people. Everyone is carrying computing devices and these form a shared workspace of documents relevant to the meeting; perhaps each has a Web server. How is this information space structured? Imagine everyone has linkbases that they publish - then we have an ad hoc aggregation of linkbases to assist in constructing the hyperstructure for the shared information space. Looking at the agenda of the meeting, you now have links available according to the linkbases of everyone in the room. Even if you do not use these directly, your agent may wish to - perhaps, trivially, for search purposes, or in general to answer a context-sensitive query.
We are developing a notion that the hypermedia middleware - the link service - needs to support ad hoc networking by supporting ad hoc combinations of linkbases. But first it must also support mobility, sharing and mobile access to linkbases. These then are the goals in our next iteration of the architecture of the Distributed Link Service (DLS). Our approaches include the use of directory services [2] and mobile agent technology [3].
We have built a system comprising a modified DLS implementation and an application which generates the linkbases automatically in the ad hoc setting. Users submit documents they deem relevant to the meeting via a shared resource; the application then extracts linking information which is made available to the DLS as a custom linkbase for the meeting. Users at the meeting set their web browser clients to use the DLS as their web proxy; as a result, documents they browse throughout the meeting have the newly created links inserted where relevant. The proxy is also able to record the users' trails and we intend these to be a source of further navigation support in later systems.
Extraction of link information from HTML documents is a straightforward task, and when working with open hypermedia systems the separate link databases are already available. The link extraction mechanism used in this prototype is a tool that parses a set of documents, saving link data in a relational database and making local copies of the documents with the anchors (HREFs) rewritten. The SQL database facilitates the subsequent generation of multiple linkbases. The current implementation is based on libwww from W3C and runs on FreeBSD and Linux.
The purpose of the prototype system is to discover whether the DLS can facilitate dynamic ad hoc linkbase creation to enhance navigation of the pervasive information fabric. We have shown that this is a feasible goal, and in the process we have identified points for further research. One is the extent of the ad hoc link space structure both space and in time; e.g. to what extent local and global resources are used, and the associated security and authentication issues. Another issue is the policies concerning retention of linkbases when the meeting is over and the fabric discarded; e.g. how can the meeting be retained in an archival form for customised retrospective navigation?