Chapter 11. Conclusions
11.3.2. Personal reflections
• The current systems supports text, audio and 3D graphical communication but not real-time video. This is an area of ongoing work being explored by Gail Reynard in the Communications Research Group as part of her PhD studies.
• In normal working situations it will be appropriate to integrate the CVE system with the other tools which the user makes regular use of such as World Wide Web browsers, word processors and email facilities. This is not currently supported by either system.
11.3.2. Personal reflections
This section gives brief personal reflections on the work presented in this thesis and the activities which lie behind it. The aspects considered in turn are: “how it went”, “how it turned out” and “what should have been done differently”.
How it went
I (the author) think that the work presented here started particularly well, with a first version of MASSIVE-1 being usable about six months into the period of study. This was important because the overhead and time requirements for obtaining useful expe- rience with the system were unexpectedly demanding. The timely occurrence of the BT/JISC funded Inhabiting The Web project allowed the system to be used to hold 17 wide-area meetings and provided extremely useful experience backed up by activity and traffic logs.
A large amount of time was consumed by aborted developments of intermediate ver- sions of the successor to MASSIVE-1. I was unable to locate a distribution platform with which I felt happy with and which was appropriate to the structure of the second system. This was also complicated by the relatively late completion of the network analysis of section 6.3.6 which demonstrated the importance of network supported multicasting for reducing total network bandwidth requirements. With a target of 100 simultaneous users it became clear that multicasting would have to be employed. This further reduced the choice of suitable distribution systems and I decided eventually to create a new distributed object system which integrated multicasting in a natural and relatively light-weight fashion.
In parallel with the development of the core system facilities I was also reflecting on how the constraints of multicasting (i.e. common delivery to a set of observers) might be reflected in an awareness model and vice versa. Consideration of objects as repre- sentative of multicast groups was the basis from which the third party object concept was refined and developed during the first half of the third year of study. The comple- tion of MASSIVE-2 was pushed back towards the end of the third year of study, and this was exacerbated by beginning to lecture three months into the third year.
There has not yet been time to exercise MASSIVE-2 in the way that MASSIVE-1 was in the ITW trials. However the NOWninety6 poetry performance using MASSIVE-2 (see section 9.2.1) was a significant motivation (understatement) to get the system operational, reliable and user-friendly. My effort has been diluted over the final year of study by teaching and other research commitments and I have concentrated on incremental improvements to the system, together with the evolution of the analysis and evaluation presented in chapters 9 and 10 in particular.
11.3.2. Personal reflections
My working style tends to be one of rapid prototyping and exploration in parallel with conceptual work, modelling and analysis of existing systems. This is apparent in the evolution of the third party object concept alongside the development of MASSIVE-2. I find that this can give a richer (and more interesting) development process with a more realistic “dialogue” between theory and experience.
How it turned out
MASSIVE-1 was explicitly a minimal prototype implementation of the spatial model of interaction. As such I think that it has been very successful. It has provided a dem- onstration platform to explore the concepts of the original model. It has also been use- ful for holding distributed meetings over wide area (though high-bandwidth) networks. It has been particularly satisfying to see others deriving new insight from its use and enjoying involvement in virtual meetings.
MASSIVE-2 was a more ambitious system intended to be a general CVE and applica- tion development platform. I consider it to have been partially successful. For exam- ple, it successfully hosted the NOWninety6 poetry performance. As an example of a different class of application the WWW-3D collaborative web-browser (see section 9.2.4) has been ported to MASSIVE-2 from the DIVE system [Hagsand, 1996] and region-based enhancements have been made to it. Other members of the Communica- tions Research Group at the University of Nottingham are also developing their own work with and within the system. However, preliminary performance tests suggest that the system will support approximately twenty mutually aware users on typical current workstations (e.g. Silicon Graphics O2); this is still somewhat short of the design goal of 100! Likely reasons for this appear to be: the relatively heavyweight character of the distributed object system, especially when handling continuous media (e.g. audio); and the current focus on peer-to-peer communication, with little facility to off-load or share computation with other machines in the network. The support for abstractions should address this second issue to some extent, but I think that it will need more dynamic and flexible management than exists at present.
I am very happy with the third party object concept. In particular, it has integrated with the original spatial model of interaction in a very elegant way. It has also demon- strated at least some of its flexibility in MASSIVE-2. The notion of using (negotiated) awareness to control the third party object appears particularly promising and expres- sive. MASSIVE-2 is currently rather limited in this respect, concentrating as it does on spatial membership controlled third party objects (i.e. regions), however there appears to be significant unrealised potential within the model that might be used in future systems.
What should have been done differently
I regret the time which was effectively wasted on aborted versions of MASSIVE in the middle of this work. I would ascribe this to a lack of early modelling and perform- ance evaluation, and also to a general tendency to start prototyping a little too soon. In principle and in retrospect I would like to have found an existing distributed system platform to use, or possibly used an existing virtual reality system or toolkit such as DIVE or MRToolkit [Shaw et al., 1992]. This should have allowed me to concentrate more on the individual issues and areas which I was exploring. However, I doubt that this would ever have happened. The areas being explored deal with relatively deep