Yet another flamebait comment
Ted is pretty close to dead-on when it comes to the Public Sector on this. Government doesn't work the way most does; in fact, when it comes to software licensing, they're buying it in such massive quantities that they get the deepest discount available.
In our corporate environment, we pay about $200USD annually for our Microsoft EA. The EA includes XP/Vista, Office 200x Professional. $600 every three years for products that everyone is familiar with, that all of our clients and vendors use to send data back and forth, as well as integration with our email, document management, instant messaging, service desk, software distribution, software patching, directory services, and CRM suites. For 1,000 employees making an average of $25/hr who are able to bill for $50/hr, it takes about 8 hours per year to recoup that cost.
When comparing the costs of F/OSS with closed source software, there is absolutely no way that I can start relationships with new vendors, upgrade/migrate/transition all of my thousands of desktops and servers, enforce global standards for desktop management, leverage enterprise-wide 802.1X from a single interface, or any one of a hundred other technology and business tasks I need to do to keep running. I would have to coordinate with my clients to ensure that any software macros or custom application scripting would work, thereby increasing development time. I have about 140 custom-developed applications that would need to be updated and regression tested. I would need an army of well-trained staff to teach my "I only use Windows XP and Microsoft Works at home" how to use OOO, and then another army of service desk folks who were fluent in both Microsoft Office and OOO to be able to help users with branfarts remember how to do tasks the new way.
Then I have to worry about all of my purchased and customized LOB software, integration with special hardware, migrating document management and email systems, supporting mobile devices, and other special hardware with integrated desktop components (integrated scanner/copier/fax devices, especially).
At that point, all F/OSS has done for me is cause a lot of headaches with cost me a fortune in both hard and soft costs for training and application/data migration. It would (conservatively) take an average of 8 hours of data migration and PC refresh per employee, 40 hours of training on new desktop productivity, ERP, financial, sales, and other applications, and productivity reduced to about 70% for 6 months while they become proficient on the replacement applications. We're looking at 160 hours per employee for lost time, which will hopefully be a one-time hit. If I'm making $25/hr net per employee on my 1,000 employees, that's $4m that I have to figure out how to recoup. My ramp-up software costs for new employees will be minimal, but my training and lower-productivity costs will be higher.
Factor in the IT time (if I contract it out at $50/hr for the desktop refresh/data migration parts, and I need 8 hours per desktop/user, that's another $400,000 in labor), and my cost is $4.4m to change out my closed-source stack to an all F/OSS stack. Versus my $200 per year per employee to maintain my current licensing, it's over 20 years before I break even.
With today's economy, we might not even make it 2 or 3 years. The cost argument for F/OSS just doesn't hold water.
And we haven't even talked about server applications, or any application redevelopment that needs to be done, nor have we discussed how I'm going to administer this environment. Additional training or hiring of IT staff and consulting services, etc. And I still probably want to maintain some level of vendor software support, because it's impossible for any staff to have all the answers for all products in the environment.
In the public sector, where the scale of users and applications is multiplied by thousands, there's no way it's economically feasible. Even if all of the software licensing, implementation, data migration, and support costs were "free - as in beer," there's still potentially millions of man-hours wasted for upgrades, re-training, and troubleshooting with a whole new set of tools.
Government is already horribly inefficient as it is; how would you justify to the taxpayer even 20 lost/wasted/less productive man hours per year for the US Federal Government's 920,000 employees?
While F/OSS products may be better, more secure, more stable, or any of the other flame-bait arguments that get used to support it, the financials just can't support it yet in the public sector.
Mine's the one with the ROI calculator in the pocket.