Re: [Fsfe-uk] Data on UK Linux usage

[Alex Hudson]

On Thu, 2003-11-20 at 00:25, MJ Ray wrote:
> > Bit of lateral thinking -- why not get stats from web sites thn can 
> > reliably differentiate the web client used and measure the change in 
> > Linux-based clients?
> 
> Given the numbers faking or blocking User-Agent headers (which are 
> higher for the systems of interest), how do they do it reliably?

You can actually do it fairly reliably - while it is completely
impossible to do 100% accurately, a combination of user-agent with a
backup Javascript check will generally tell you what the browser is
(think along the lines of how nmap works). 

I'm not sure that data tells you what the original poster wants anyway -
the GNU/Linux deployment in this country is unlikely to be desktop-based
in any large proportion, so I would expect the number of web browsers to
be in use to be fairly small. The number of people with (say) IPCop at
home, but a central Windows desktop is quite big. 

I don't think there is any way of getting a reasonable number of OS
deployments, but then, that's not the case for any OS - even Microsoft
are unaware of how many people use Windows in this country (since that
probably gets shared as much if not more than GNU/Linux...). Best-guess
estimates would probably have to be based on the number of units shipped
(since we have a fair idea of how much hardware is out there, and what
rate it gets replaced and grows at), and perhaps some of the main distro
mirrors have country-based stats on who downloads copies? We're still
talking very much about probablistic methods, though.

Cheers,

Alex.