The ROADS project exit strategy - Ensuring the future of ROADS for its users

The eLib funding for the ROADS project runs out at the end of July 1999 but that doesn't mean the software will disappear or its users will be left holding the baby. Paul Hollands outlines the plans for the future of the ROADS project and the code-base explaining how the team hope to ensure the use and development of ROADS well into the new millennium.

roads.opensource.ac.uk - ROADS into the next millennium

The end of eLib funding

The current round of funding for ROADS was provided to keep the project in a holding pattern until the Resource Discovery Network comes into being in August 1999. One of the requirements of the eLib programme is that all of the funded projects develop a suitable exit strategy. Another of the main objectives was to disseminate project outcomes to the rest of the community.

In practice both of these issues have been tricky and there have certainly been some eLib projects which have produced software tools and other outcomes which for one reason or another have disappeared at the end of the project's funding and been lost to the wider community. For a long time there was no clear mechanism to feed their results back to the wider community in such a way that would allow practical take up by others.

In contrast ROADS has been highly successful in this regard and has been so well beyond its original remit. ROADS was originally funded to provide a set of tools to the ANR eLib subject gateways. A number of these gateways have used ROADS and continue to do so. Because of our participation in the DESIRE programme there has been huge take up of ROADS in Europe and particularly in Scandinavia, and we now have a large number of well established gateways who are not only using our software but who are contributing to its ongoing development.

Since the release of ROADS version 2 and the arrival of the code-base on CPAN our downloads of version 2 alone are up to one and a half thousand. (This is the figure from our servers. We have no idea how many other people have downloaded ROADS from the Metalab Linux Archive or the 250 CPAN mirrors around the world.)

It is clear that it is now time for ROADS to move beyond an eLib project. ROADS needs to have a secure future independent of JISC / eLib funding, even if the project gains more of this funding in the future.

As it stands the project has run far longer than anyone initially anticipated. By the end of July this year the code-base will be available in version 3. This iteration will be much more modular than previous versions and allow much more flexibility in the way the software can be implemented. It will also mean that development of ROADS can be much more distributed with users contributing improvements and extensions back into the code-base more easily.

Doing ROADS without the money

The Resource Discovery Network Centre has expressed an interest in funding ROADS support for its member gateways for a short period while its technical review is underway. The RDNC has a development budget which will be allocated following the technical review. It is possible ROADS will get some of this money, but this funding is in no way guaranteed in the long term. Also, even if the RDNC were to offer some funding, they cannot support R&D in the same way that eLib did.

So at current reckoning our future funding will be at best minimal but more probably non-existent.

It may seem that it will be impossible to keep ROADS going without the funding from eLib and this is certainly true of the way the project is run at the moment. We have therefore investigated models which we could move to which would enable us to continue to develop the ROADS code-base and ensure its availability at marginal cost to us and preferably free for users.

The best model that we have come up with is that of the various Open Source projects which have been springing up in the wake of the recent explosion of interest in Linux.

The ROADS project has in fact been doing Open Source long before the term was coined. ROADS has always been distributed under the GPL license, the original free software license.

ROADS would not have existed without what has come to be called Open Source software. ROADS was developed using free tools such as Perl, Apache and Linux.

The eLib programme has benefited in no small measure from this approach as we have been able to leverage the expertise of a large core of ROADS users who were technically outside our remit to support and this has meant faster development and a much better ROADS tool-kit than would otherwise have been the case.

The aspect of the Open Source model which seems most relevant to ROADS is this Open development approach whereby the software is written and its development coordinated by its committed technical users. We actually do this already, but we would like to extend the current model we are using to make it easier for more users and developers to contribute while avoiding the problem of 'balkanization' of the code-base.

Jon Knight and Martin Hamilton, the two original authors of ROADS have made a commitment to continue to maintain the ROADS code on a best effort basis for as long as others continue to use it. Jon and Martin will also act as benign dictators over decisions about what is included and excluded in future versions of ROADS. This model has worked very well for the development of Linux.

If at any point they are no longer able to do this, by using the Open Source approach we can ensure a smooth transition to new custodians for the ROADS code. The ILRT also has a vested interest in ensuring ROADS continues to be maintained and developed as we have a large number of services based here which rely on the software. For this reason the ILRT has agreed we will host the new ROADS Web site on our servers under the banner of a new domain opensource.ac.uk

We will also continue to submit our own ROADS developments for their inclusion back into the code-base.

Finally, the ROADS partners will continue to work together on a new but related project to develop an IMESH tool-kit.

The benefits of the Open Source model

There are a number of benefits to developing software under the Open Source model which have already greatly increased the successful take-up of the ROADS software:

It is hoped we can capitalize on these benefits even more in future to ensure the ongoing development of the tool-kit.

users can become developers thus development becomes user focused
bugs are fixed more quickly as there are more people looking at the code and submitting fixes
bad design doesn't linger, it is developed out too quickly

A side effect of the Open Source approach is that it would create what is essentially a free-market for support. People who require commercial level support could pay us (or anyone else) in return for guaranteed response times etc. It is probably too early to able to tell whether this is appropriate for ROADS (we would be keen to hear your opinions on the open-roads.net.lut.ac.uk list) but if it worked this could provide an income stream for the project.

The ROADS Clearing House

The new ROADS development strategy will be focused on a code-base 'Clearing House' on the Web. This site will follow the model of a number of other Open Source projects such as Mozilla.Org, the GNOME desktop development project, and the more venerable Apache project.

The development of the ROADS code will most likely be handled using a tool such as CVS (Concurrent Versioning System) which allows developers to monitor and synchronize with each change other developers have made as they happen. We will also probably take another leaf out of the Linux development book and assign responsibility for different sections of the ROADS tool-kit to a number of individuals who would then oversee the development of that section. Jon and Martin would remain final arbiters of what was included or excluded however.

Further information about how to contribute to Open Source projects via CVS is available on the GNOME site.

We already have the necessary mailing lists in place to continue to support users on a best effort basis and we are currently investigating software for tracking bugs such as the GNATS system used by the Apache group.

A possible manifesto for the ROADS Clearing House?

We will:

collect changes and help authors synchronize their work using CVS, and periodically make new source releases which incorporate the best work
operate discussion fora (mailing lists, newsgroups, or whatever seems most appropriate)
coordinate bug lists, track and publicize works in progress, and generally attempt to provide "road maps" to the code, and to projects based on the code
break out generally useful code into independent library modules and post to CPAN
be code integrators, and, through our fora, try to help people reach consensus