post-AstroGrid

Tony Linde
AstroGrid Project Manager
University of Leicester

Introduction

When a new way of doing things becomes available, especially when it develops as quickly as the VO is developing, many people's first instincts are to try and take control of it; to put in place organisations, structures, rules and processes to make sure it is done properly, to stop it getting out of hand. It is this instinct which I think we ought to challenge.

A significant part of AstroGrid's funding goes towards satisfying various committees or other bureaucratic structures: AGLI, AGOC, GSC to name the main three. While I am sure each body would argue its essential purpose, I would estimate that at least £50K of AstroGrid salary spend goes towards these purposes each year. Adding more such committees, oversight bodies, and other structures to UK astronomy is not, to my mind, a good use of funds.

I have worked for many IT teams within many types of organisation. In every case where a small, focussed team has been set up to address a single technology or market niche (desktop applications, web-based applications, middle office processing etc.), within a few years, that team will have acquired several layers of management, rules, processes, standards and will have sought to expand its workforce and sphere of influence. This will happen despite the original intentions to keep it small and focussed.

I don't intend to minimise the need for standards, but to argue instead for minimal standards. The more standards that are imposed and the more bureaucratic the processes surrounding them, the longer it takes to get changes to those standards agreed or new standards set. We see this especially in the web arena where companies like Microsoft approach standards bodies after they have established their own version of how technology should work.

The rest of this document presents my own personal view of the post-AstroGrid, IVO world, recommending minimal standards, project-based development and the leanest possible organisation.

VO philosophy

I would like to see the VO develop akin to the web. The web exists, not as a single entity, but as millions of systems - hardware, software, services and data - which basically implement a couple of simple standards: the http protocol and the html language. Yes, there are supporting protocols like tcp/ip and dns, and incremental standards like css and xhtml, but the style and import of these standards allow as many people as possible to connect to the web with whatever resources are at their disposal.

In the early days of the web, Microsoft created MSN as its own web: you had to pay to subscribe and could only see MSN-based pages if you did subscribe. They tried to create a single portal to the web with control over who placed content on their web and what users could and could not see - they failed. I would hope that anyone who tries to enforce single entry points on the VO will likewise fail.

My view of the VO is that a few simple interoperability standards should serve to allow anyone to build components, portals and client software which can make use of any other resource on the VO. One group might build a portal, with their own branding, offering perhaps a customised news viewer, but with all the querying, job control and visualisation components derived from services elsewhere on the VO. Another group might simply package a data analysis module as a web service and make it available on the VO for others to incorporate into workflows.

The goal of this approach is to make it easy for people to add new data and services to the VO, ones which improve on what has been built before, with no need to gain some controlling organisation's approval before doing so.

Organisation

What is the absolute minimum organisation that is needed to sustain the VO in the UK: none, in my opinion. However, there are areas in which an organisation would prove useful. Having one body in the UK to speak on VO matters would be useful. Peter Quinn, of ESO, has already said that he would like to see only national bodies interacting, rather than every data centre and research group in Europe.

However, instead of creating this body, staffing it and imposing it on UK astronomy, it should be created from the data centres and research groups. So, people should be delegated to represent their groups in this body with elected officials to represent the country at European and international level. Funding bodies will need to recognise that those who take on the official roles will do less work in their own institutes, but I'm sure models exist for this already.

As I've said, this body should represent and coordinate the UK interests, not dictate them. While AstroGrid personnel may fill some of these posts initially, that should not be a given nor a pre-condition.

Technology

What is the absolute minimum organisation that is needed to sustain the development of the VO in the UK: none, in my opinion. And I don't wish to qualify this answer. As I have said above, creating any sort of IT team will ensure that it grows and becomes bureacratic: that is inevitable; it is the nature of any organisation.

What I would like to suggest is that any technology development be project-based. That is, if anyone wants to add a function to the VO, they should propose a project and get funding for that project. Likewise, if a science project can extend what they do and add it to the VO, they should be able to apply for funding to add this capability to their project.

It is likely that there will still be a significant amount of work to do in building additional VO functionality after the current AstroGrid project. At the very least, we need to continue our work on integrating new standards into our components and making it more 'griddish'. I'd suggest that we should propose AstroGrid-II, probably over two years and with half the number of developers.

Standards

One of the main goals of the organisation I mentioned above will be to represent UK interests on the standards setting bodies, particularly within the IVOA, since this seems to be the natural VO forum at the moment. We must ensure that our representatives are drawn from both the data centres and the research bodies to ensure the VO develops into something beneficial to both.

I would also argue for minimal standards. What we want to achieve is the ability for components to work together, not to force everyone to do the same thing. Some of the things I've heard in the various forums indicate a push towards a single (or national/regional) VO system. We should use the standards bodies to ensure a vision of the VO that enables anyone to create software, add services and data and use the VO.

Standards should not enforce the way a component works, but the way components interact. So, our registry could be built one way with fine-grained detail and someone else's registry be coarse-grained and built in an entirely different way. As long as each can enquire of the resources held in the other then that is as far as the standards should go. The same applies to all other areas of the VO.

Finally, the standard should define a basic mode of interoperation; a mode which is relatively easy for all to implement. Suppliers should be free to extend the standards and offer enhanced services (eg more complex and interactive queries of the registry, or ontology-based queries). For this they will offer enhanced interfaces. They must also support the basic interface defined by the standards but the ability and freedom to offer enhanced services will drive the evolution and rapid development of the VO. These enhanced methods will eventually become the basic standard.

There is no way of making a standards body responsive and dynamic; it is the nature of the job that requires caution and much deliberation. What we want is to ensure that development is not hindered by the standards.

Conclusion

I guess the conclusion is that I'd like the AGLI to develop a view of the post-AstroGrid scene that is non-bureaucratic, dynamic and open to all. I hope that the above contributes to this discussion.

-- TonyLinde - 20 Jan 2003