Birmingham City Council has defended its year-long trial of desktop Linux, claiming it to be a success, despite an independent report showing it would have been cheaper to install Windows XP.
The ups and downs of a year-long rollout
In an exclusive interview with PC Advisor's sister website Techworld, Glyn Evans, head of IT for the council, argued that the higher cost resulted from the council having to experiment with the new technology and build up a depth of technical understanding, as well as fit it with the complex system already in place.
The £105,000 saving that the report says would have resulted from going with Windows XP has also come under question, as it was calculated using the special discounted licence rate that Microsoft offers councils – something critics argue is a calculated effort to prevent public bodies from building up technical knowledge of open-source offerings.
With Birmingham's trial period over and with lessons learnt and understanding gained, the council now expects to make cost savings over time, and contrary to press reports that claimed Birmingham had scrapped the Linux initiative, it will in fact "significantly increase" its use of open-source software, Evans said. The trial also had other positive results, he claimed, such as demonstrating the ease with which Firefox and OpenOffice.org can be substituted for Internet Explorer and Microsoft Office.
The trial was carried out with the government-backed OSA (Open Source Academy), and planned to install Linux on 330 desktops in the council's libraries service, split between staff PCs and public access terminals, in an effort to build up practical experience that could be drawn on by other public-sector bodies.
It ran from April 2005 to March 2006, but is still ongoing, with the council refining its Linux desktop image and planning further rollouts next year, according to Evans. "The project did not end when the element of original funding ended, because it is part of the Library Service strategy," he told Techworld. "This project is still very much ongoing, and now that a stable image has been developed, we would expect significant movement forward."
He admitted the council's original plans were over-ambitious, with rollouts of Linux-based staff and public PCs originally scheduled during the one-year trial period. In reality, ongoing testing of the desktop configuration means no Linux desktops have yet been installed. Instead, 96 public desktops and 134 staff desktops are running open-source applications such as the OpenOffice.org office suite and the Firefox browser.
The council does plan to begin migrating those desktops to its Suse Professional 9.3-based desktop OS (operating system), however – a plan that should go into action in the near future, according to Evans. He said that far from scrapping the Linux initiative, as has occurred in some other high-profile cases such as the London borough of Newham, Birmingham is planning to "significantly increase" the number of desktops involved with the project.
Evans' description of the project is in sharp contrast to the findings described in a case study authored by iMpower Consulting at the formal conclusion of the trial in March, which is available from the OSA's website (a pdf of the report can be found here). The case study found that the council had failed to make a business case for its Linux desktops, largely because the £500,000 cost of designing and implementing the system cost more than the estimated cost for a Windows XP installation.
The difference is largely down to high "team costs", including setting up the project, technical definition and design, development and testing and training, all of which amounted to roughly £100,000 more than the estimated team costs for a Windows installation. The total cost of the trial was £534,710, compared with an estimated £429,960 for Windows XP.
"The project showed that there are considerable costs incurred in decision-making, because of the huge range of open-source options available," said iMpower in the case study. "The extra resources involved in decision making and project management mean that the cost of this first-time open source implementation for Birmingham City Council was significantly higher than for a comparable proprietary upgrade, despite the minimal licence costs for open-source software."
The case study also detailed the many frustrations involved in approaching an unfamiliar desktop technology, including the discovery that key applications wouldn't run on Linux and usability problems with the original Gnome interface. At one point, realising that most of the usability issues were attributable to Gnome, which had taken three months to configure, staff ripped out Gnome and replaced it with KDE. The new interface was up and running within a week.
But evaluating the project solely on what occurred during the original trial period is absurd, according to Evans. "There is no doubt that startup costs for this project would be high because of the level of requirement, the level of Linux expertise within Birmingham City Council and the complex requirements of the library service for the public desktop," he said. "The positives centre on future costs."
Birmingham's requirements involve "much more than purely tweaking a standard desktop image", according to Evans, including the need for particular security features, authenticated processes and the supply of specialised management information for performance monitoring requirements.
For instance, existing Windows 3.1 public terminals used a program called Deepfreeze that rebooted the system at the end of each session, something that had to be re-engineered for Linux. Another problem arose with the handling of removable media, which often wasn't recognised or caused errors on the desktop.
Staff also found that the OS was storing information about the contents of public users' removable media, and for privacy purposes had to develop a script to delete this information, which caused further delays in developing the final image.
Linux simply wasn't able to meet certain requirements, such as the ability to run Galaxy, the library management system. The council couldn't afford to pay Galaxy's developers to port it to Linux, and running it in emulation would have added yet another layer of complexity, so many staff PCs were simply migrated to Windows XP with OpenOffice and Firefox.
All this planning and configuration added to Birmingham's startup costs, and meanwhile, the fact that Birmingham qualifies for Windows discounts further lowered the comparative cost of a Windows installation.
The council gets a steep discount on Windows licences through a broader Education Select licence arrangement, paying £58 for a Windows XP licence compared to roughly £100 for OEMs (original equipment manufacturers). "Accounting for corporate instead of Education Select licences would have added nearly £50,000 to a Windows upgrade project," iMpower found.
Despite this, however, the council feels that further down the line the investment in open source will pay off – for instance, Linux-based systems can be upgraded incrementally, avoiding large one-off licence payments as would be the case with a Windows upgrade. Any number of further desktops can be added to the project without adding extra licence costs.
Graham Taylor of OFE (Open Forum Europe) said one of the key concerns emerging out of the trial is the effect of vendor lock-in, with particular key applications dictating the choice of OS. "For me, this was the major issue emerging," Taylor said. "Our estimate is that up to 90 percent of UK public-sector organisations have this as their current position, and can no longer freely choose next steps in procurement."
OFE and OSA have developed a certification scheme called Certified Open, designed to encourage applications to certify on open-source platforms, which will launch in the new year. While attempting to design and implement a Linux desktop system that could be used by staff and the general public with limited technical knowledge turned out to be an onerous and frustrating chore, by contrast, many of the open-source applications themselves ran smoothly and went over well with users.
OpenOffice, for one, met little or no resistance with most users, many of whom said they didn't notice they'd been using a different application. (Power users did face some problems.) The public had no trouble using Firefox on public terminals and some said they preferred the open-source desktop to Windows. "It appears that OpenOffice provides a satisfactory equivalent to Microsoft products for those using basic or intermediate functionality," iMpower found.
The trial's findings will be used by the OSA to give other public-sector bodies background when they consider using open source. The OSA has backed other, more unambiguously successful open-source projects, such as Bristol's implementation of StarOffice, which saved it hundreds of thousands of pounds in one-off licence costs.
The UK has less than average usage of open source compared with other EU countries, according to a report by the University of Maastricht, with 32.1 percent of all UK local government users on open source compared to the 78.7 percent European average.
That lack of experience adds to the difficulty of public sector bodies getting involved with open source, iMpower found. One high-profile open source failure was the London borough of Newham's decision to scrap an open source trial in favour of upgrading to Windows XP in 2004. That came following Microsoft's offer to provide free consultancy to the council and a subsequent deal struck with Newham Council that remains undisclosed but which is widely assumed to offer a huge discount on Windows licences. Newham Council will be appearing alongside Microsoft today at the launch of Windows Vista and Microsoft Office 2007.
In October 2003, the OGC (Office of Government Commerce) and the OeE (Office of the e-Envoy) announced they would fund IBM to run nine proof-of-concept open-source trials designed to measure the cost benefits of open source over proprietary software such as Windows. Participants were to include the OeE, the Office of the Deputy Prime Minister, Powys Council, Newham and Ofwat, with Newham dropping out. None of the trials have led to further rollouts.