One of the interesting omissions from the Microsoft case study cited yesterday is any support for the claim that the Wintel decision will save the state $10.5 million over five years.
It's the headline number and one we're not supposed to question - but put it in the context of 57,000 PCs over five years and it works out to 15.3 cents per working day - not exactly a lot, and but still something that we should be able to check up on.
In fact, however, I wasn't able to find anything - indeed Microsoft fails to either raise or address three distinct groups of questions about this number:
- how was it derived?
- who's $10.5 million is it? and,
- how does that estimate compare to the savings that would have been cited had other people done comparable analyses in the context of alternative technical options?
We have no information on group one - I could find no published analysis whose assumptions or results we can ask questions about and no publically available audit results to show how realistic this number has turned out to be.
We do have some information about group two: whose dollars these are. Ultimately, of course, these are tax payer dollars - but the situation summary in the Microsoft report tells us about some of the intermediate players. Specifically it says:
In the past, each agency was responsible for its own hardware and software procurement, implementation, and management, resulting in a vast number of disparate operating systems and applications. The state opted to consolidate all the agencies's IT services into a single entity called the Bureau of Communication and Computer Services (BCCS) in the Department of Central Management Services (CMS)
The very best best assumption you can make, therefore, is that the BCCS five year plan for the affected services is $10.5 million less than the total they got by asking each agency about current spending on desktop and messaging services. Unfortunately it's a sure bet that not a single one of the agencies involved would have had either a clear five year cost projection for the status quo or any motivation to share accurate numbers with an imposed central authority.
In the context of that decisive $10.5 million dollar savings it's important to remember, therefore, that a guess made up from multiple guesses, all of them self-serving, isn't any better than a guess made up, by the group most likely to benefit from its acceptance, as big enough to matter but too small to challenge.
The report says they decided against things - against IBM because they saw Lotus Smartsuite as less integrated than Exchange and Office, against GroupWise because Novell was going Linux, and against Linux because it's just a "science project" - but there's no information about the third set of cost issues: how did the cost of these alternatives compare?
On that, however, we can use external information to make some guesses. For example, because the list price for Lotus smartsuite for Windows was about one third that for the Microsoft product set selected and IBM's policy at the time was to reduce the price to meet customer expectations, it seems highly unlikely that BCCS could have mounted a capital cost argument for their decision - and because IBM has since moved the product to Linux any argument based on operating cost would have evaporated too.
Since you can do the same with the Novell option, the bottom line is that there's no possible way to get either capital or long term operating costs for the Microsoft solution below those of its alternatives. The $10.5 million must, therefore, have come from inputed user productivity gains - apparently worth 15.3 cents more per user day than the extra cost of buying and supporting Microsoft.
Think outside the box, furthermore, and the fact that they don't disclose anything about their cost structure while listing only traditional client-server options for rejection, suggests that they probably didn't look at migrating to Solaris with Sun Ray desktops.
Had they done so, however, they could have used Windows servers to deliver all of the software they appear to value, while completely avoiding the forced march to Vista, dual core processors on every desktop, and next year's Windows Server 2008 incompatibilities -all of which is going to cost the Illinois taxpayer an average of about $1,800 per user -roughly $100 million new Wintel dollars in total.
And if that's not enough - there are ginzu knives in the offing, or, at least, another beautifully written put down of the whole Windows imposition by BCCS. This one is by another unknown writer contributing to Sun's Java desktop system write-up:
Designed to thrive in a Windows-centric world, Java Desktop System interoperates with Microsoft Office files, networked Windows printers, and Exchange servers as well as standards-based LDAP, mail, calendar, and Web servers.
Basically, if Illinois had stayed with Lotus or GroupWise they'd now be transitioning to Linux and either SmartSuite or OpenOffice; if they'd evaluated and selected Solaris 10 with Sun Ray desktops they could have kept all their Windows software except the desktop OSes and now be transitioning smoothly and at virtually no cost to standards-based products like OpenOffice, Lotus Domino, and Sun's java desktop system.
Know what I think they're doing instead? Evaluating hardware to run Vista, fighting a losing battle on control and security, re-inventing mainframe era administrative controls to keep those state agencies in line, and figuring out how to justify their next hundred million in hardware and software upgrades.