The Billion Dollar Kernel
jesgar writes "The Linux kernel would cost more than one billion EUR (about 1.4 billion USD) to develop in the European Union. This is the estimate made by researchers from the University of Oviedo (PPT), whereby the value annually added to this product was about 100 million EUR between 2005 and 2007 and 225 million EUR in 2008. The estimated 2008 result is comparable to 4% and 12% of Microsoft's and Google's R&D expenses on whole company products. Cost model 'Intermediate COCOMO81' is used according to parametric estimations by David Wheeler. An average annual base salary for a developer of 31,040 EUR was estimated from the EUROSTAT. Previously, similar works had been done by several authors estimating Red Hat, Debian, and Fedora distributions. The cost estimation is not of itself important, but it is an important means to an end: that commons-based innovation must receive a higher level of official recognition that would set it as an alternative to decision-makers. Ideally, legal and regulatory frameworks must allow companies participating on commons-based R&D to generate intangible assets for their contribution to successful projects. Otherwise, expenses must have an equitable tax treatment as a donation to social welfare."
That's assuming that a replacement would be 12m lines of code. I recently rewrote a few classes for an open source project that I contribute to and replaced 5,000 lines of code with 500 (which did more, ran faster, and fixed some bugs along the way). Just because the current implementation is 12m lines, doesn't mean that the correct implementation is 12m lines. From the Linux kernel code that I've read, I suspect that there is a lot of redundant and duplicated code in the kernel. I wouldn't be at all surprised if you could implement it with a cleaner design in closer to 1m lines of code.
I am TheRaven on Soylent News