I wrote in an October 19 blogpost that HP (HPQ), IBM (IBM), etc. “created” Linux and followed up on October 22 concerning the Linux Foundation (LF) approach to explaining the same statement. LF, which is funded by HP, IBM and so forth had estimated (see its 12-page document here) that the R&D investment in time and money represented in a major Linux distribution… had the work all been done:
- by one company
- using a closed-source development methodology
- from scratch
would have been $10.8 billion.This amount applies specifically to Fedora 9. Six years of effort would have been required (assuming 10,000 developers, documentation specialists, QA staffers, and so forth). Fedora is the community that provides the base code of Red Hat (RHT) Enterprise Linux.
This effort is much wider than LF. LF takes care of the Linux kernel but Fedora 9—as with many Linux distributions—includes not only the relatively small Linux kernel but many utilities and supporting pieces of software. Examples include the GNU utilities and X Window, among many other pieces of software (for example, the Mozilla Firefox browser and OpenOffice.org).
The LF numbers raise the interesting IT investment research question:
"What does a major software distribution really cost (vs. this theoretical cost)?"
The best way to tell is by comparing the LF’s numbers for Linux with numbers from a software company that still takes on such major software-development efforts independently, with a closed source development model. No one takes on such a project from scratch so a perfect comparison is not possible. There is basically only one such comparison point, Microsoft (MSFT).
Over the period roughly comparable with the development of Linux as defined by LF in its study, Microsoft was developing its Vista client operating software, Windows Server 2007 server software, Office 2007, Internet Explorer 7 and other related software. Microsoft spent around $14 billion on that effort by my estimate (see illustration; sum of "Est. Comparative R&D" ) making the LF estimate seem reasonable.
Whether the two product sets are exactly comparable or whether my estimation technique is perfect would make an interesting debate but I was simply looking for reasonableness. My source is Microsoft SEC filings for the years during which Fedora 9 was developed. I took the overall R&D spending of Microsoft in those years, reduced the total in proportion to the revenues of the client/server/office-worker divisions during those years, and cut that number in half on the premise that those Microsoft divisions were also doing maintenance and documentation and support and other R&D activity on older client, server and office products.
If you want to assume that only 33% of Microsoft's R&D expense in those divisions was directly related to Vista, etc., then the LF and Microsoft numbers line up almost exactly. If you want to assume that 67% of Microsoft's R&D expense was devoted to Vista, etc., then the LF supporters such as HP, IBM, etc. got a hell of a deal. But either way, the LF theoretical model seems to have worked in terms of predicting what the LF ecosystem development expense would have been.
So now the $64,000 question is:
"How much does Linux's collaborative non-closed-source development methodology save HP, IBM, etc.?"
And does a significant amount of that saving drop to their bottom lines? More to follow.
(Note: dividing Microsoft's R&D numbers in proportion to the divisions' revenue does not take into account the cash-cow effect Microsoft might be employing and other ways in which Microsoft may have actually divvied up its R&D spending. It might have spent the whole $28 billion on XBox but we don't have the data to know that.)