[OT] Free software versus software idea patents

[Dan Stromberg]

On Tue, Apr 12, 2011 at 12:37 PM, Ian Kelly <ian.g.kelly at gmail.com> wrote:
> On Tue, Apr 12, 2011 at 3:22 AM, harrismh777 <harrismh777 at charter.net> wrote:
>> It is not measured in any way, and it is almost impossible to determine therefore in any accurate fashion. There
>> really are no data... what we need here is a census of sorts.
>
> There is at least one method of measuring it without resorting to
> sales figures: logging user-agent data from web browsers.  Is it
> perfectly accurate?  Of course not.  But there are a number of
> different organizations that do this, sampling hundreds of thousands
> of different websites, and they consistently report that the various
> versions of Windows have a total usage share ranging from 80% to 90%.
> That at least gives us an upper and lower bound with a great deal of
> confidence.  In the same data, Apple systems range from about 7% to
> 15%, and Linux musters a meager 1% to 3%.

This data is of course skewed a bit toward computers that people are
using web browsers on.  Also, something like LWN tends to get more
Linux users, while microsoft.com tends to get more windows users.
Also, some sites only allow IE (often for rather capricious reasons),
so some *ix users tell their browsers to pretend to be IE to get
around the restriction - and sometimes it's just easier to pretend to
be IE all the time, even when you don't really need to.  And IE runs
on Solaris, albeit in what is probably now an ancient version.

A better measure would probably come from getting various companies
managing various links of the internet backbone to set up passive
fingerprinting machines.  The cost of covering such links would
probably be minimal, and the performance impact likely zippo, but the
companies doing the work would probably not benefit much - so unless
someone like Redhat or Canonical, or even Apple step up and fund it,
it probably won't happen.

http://en.wikipedia.org/wiki/TCP/IP_stack_fingerprinting
http://lcamtuf.coredump.cx/p0f.shtml