Interesting seminar from Ian Dolphin of the Sakai Foundation on virtual research environments.
He's promised to make the slides available so I won't provide a blow by blow account that the nub of the problem is that the environments have got to consider reusability and remixability, so that tools can be added and reused easily by individual groups of researchers.
Definitely resonances there with the work I'm involved in on data reuse with ANDS and collections interoperability (which is really dataset reuse) with Project Bamboo.
And the reason is that if people are going to engage in cross institutional and cross disciplinary research they need access to data sets sitting in archives and share datasets generated.
To take a simple example: if one wished to do an analysis of early 19th century squatter settlements, such as the informal one at Ororral Valley, one might want to tie the names of these settlements to a thesaurus of aboriginal place names and a GIS system and show that settlements tended to be on grassy paddocks that were also good kangaroo hunting grounds and therefore possibly providing reason for conflict.
So it's my view is that it is not about composable environment but being able to connect data sources easily to these environments to facilitate reuse not only of tools but data, which of course means standards for both tools and data.
And as we know that quite a bit of academic work has skipped the fence with collaborations taking place on Google Docs and wikidot, and with collections of material being hosted on flickr,
we need integration as well with external tools to harvest and ingest external content.
The advantage about this is that this also provides a way of capturing scholarly output, so that people can not only co-operate on research but deposit the results electronically and make available both pre-prints and datasets for review, as well as generating researcher profiles for other purposes.
Potentially powerful, but I don't think we're there yet