Future work (or what I didn’t do)

Areas for possible future work are described below in terms of: models of awareness; user and application modelling and testing; network modelling; and system design and implementation.

Modelling awareness

• It is apparent from use of both MASSIVE-1 and MASSIVE-2 that both the original spatial model of interaction and also the third party object concept itself require explicit representation to the user within the normal interface. As the effects of focus, nimbus and third parties become more unusual (i.e. more “virtual” than

11.3.1. Future work (or what I didn’t do)

“real”) then users have difficulty in understanding why they perceive the world as they do or in inferring what they should do in order to achieve particular kinds of interaction. If an approach such as the spatial model of interaction is to be used more generally then it needs to be made sufficiently visible and malleable for “nor- mal” users (whoever they may be).

• Similarly, there is a need to make control over interaction more accessible to users (e.g. via manipulation of focus and nimbus).

• The systems and examples presented in this thesis concentrate exclusively on the management of and support for real-time (synchronous) communication and col- laboration. These ideas might be applied and generalised to asynchronous situa- tions and to the transitions between synchronous and asynchronous activity (this is one of the emphases of the Aether awareness model [Sandor et al., 1997] which is also inspired by the spatial model of awareness).

• The prototype systems described in this thesis are stand-alone tools which tend to dominate the user’s attention. To support more general patterns of use (such as the notion of always being “on-line” but to varying extents) it is necessary to address direct integration between the awareness model presented here and the other facili- ties provided by the user’s end system. There is also the need to integrate control and use of other individual and group tools within the same context.

User and application modelling and testing

• While the motivational focus of this work has been on the support of potentially large numbers of simultaneous users it has not been possible to perform large-scale trials within the temporal and resource constraints of this work. This is a clear area for ongoing work and is necessary to address the potential utility of the system designs and the network traffic model. Also, and perhaps more importantly, large-scale tests will be needed to assess the effectiveness of the spatial model, third party objects and the CVE approach and philosophy in general.

• With regard to capturing and characterising user activity and behaviour there is a need to consider: larger user populations including, for example, “expert” and “novice” users; a range of application types beyond small group tele-conferencing as considered here; a range of interface metaphors and technologies, including the use of immersive displays and tracking technologies. All of these are likely to have a profound effect on the type and pattern of user activity, which in turn impacts system and network requirements.

Network modelling

There are a number of well defined extensions to the network requirements modelling of chapter 10 which could form areas for future work. These include the following: • Consideration of non-trivial network topologies. The current model considers total

network bandwidth only, i.e. treating the network as a single shared broadcast seg- ment. The model might be extended to consider, for example, distinct LANs within a broader WAN, each with different bandwidth and/or cost characteristics.

• The model might be developed to explicitly address client-server style implemen- tations (the current model focuses on peer-to-peer communication). This may also include consideration of user access over low-bandwidth tail-links (e.g. modems).

11.3.1. Future work (or what I didn’t do)

This would go beyond the analysis of Funkhouser [1996] in considering cli- ent-server traffic as well as the inter-server traffic considered in that analysis. • The model is currently for long-term average bandwidth only. There is a clear need

to model and reason about transient effects such as possible peaks in bandwidth requirements associated with state transfer operations. This is particularly impor- tant in the context of low-bandwidth connections such as narrow WAN or modem connections and for provisioning of networks and services.

System design

• As noted above it has not been possible to test these systems with as many partici- pants as might have been hoped. From preliminary performance tests it appears that the performance of individual participant’s machine will be a key limiting fac- tor even with relatively small number of mutually aware users. There is significant future work to be done to characterise the actual limitations and to explore ways of increasing total system scalability, for example by adaptive control of interaction, or by off-loading activities (such as elements of audio mixing and management) to external shared services.

• Flow control and congestion management is largely neglected by the current sys- tems. This is an open research issue for multicast communication in general. It is complicated in this application by the use of several multicast groups in parallel, with continually shifting interests. It would also be interesting to consider the ways in which back-pressure might be represented to or applied to the individual partici- pants in a CVE so that they could reason about the observed behaviour of the sys- tem and modify their conduct accordingly.

• One possible application of CVE technologies in the long term is in the consumer market as a form of “Inhabited TV” in which viewers can “step inside” their TV set and join the audience or even the main content of the show. This is also consid- ered in section 10.3.3. This implies a need to move gracefully between very differ- ent modes of participation such as passive observer, mutually aware audience member, or central participant. The metaphors, interface technologies and system support for these varying modes of participation and the movement between them remain areas for future work.

• In a similar way the movement between synchronous or foreground use of the sys- tem and asynchronous or background use of it (as a tool for awareness) involves similar transitions between modes of participation.

System implementation

This section lists a number of obvious extensions to the implemented systems, in par- ticular MASSIVE-2 which is still being used and developed (unlike MASSIVE-1). • Communication management in MASSIVE-2 is based on disjoint worlds and on

spatial third party objects (i.e. regions). An interesting area of future work would be to extend this management framework to handle logically defined groups (e.g. artefact hierarchies or security-related groups) and also to integrate per-artefact communication management in appropriate cases. This thesis has argued that high-cost media such as video and perhaps also audio should be managed on a per-artefact basis because the cost to a participant of inaccuracy is so high. How- ever the system does not currently support this.