Home Messages Index
[Date Prev][Date Next][Thread Prev][Thread Next]
Author IndexDate IndexThread Index

Re: Firefox takes 28% market share in Europe

[snips]

On Sun, 15 Jul 2007 19:59:55 -0400, Dr. Livingston wrote:

> So today marketshare *CAN* be measured for free applications. Amazing 
> because just the other day it couldn't be measured for a free app. It either 
> couldn't be measured or it doesn't exist.

Really?  How odd.  There's no fundamental difficulty in telling, from
browser idents, about how many users of IE or FF or other browsers
actually hit your site - modulo a few oddities such as masked idents and
the like.

The only real trick there is in determining whether you're actually
examining a representative sampling or not.  For example, one might expect
a page dedicated to IE updates and add-ons isn't likely to get a
representative sampling of FF users.

Now, the usual discussion about counting noses isn't about free apps, or
about browsers... but about counting *Linux* seats.  That is a whole
different kettle of fish.

Take a common home setup for us Linux geeks.  There will be a
router/gateway and several machines behind it.  We'll tend to cache things
such as package data, so that updates require downloading one copy of the
data even though there are many machines being updated.  Half the machines
will never be exposed to the net directly; they're used internally for
other things.

So we're talking perhaps 6-8 machines, of which maybe two ever fire up a
web browser, and *those* go through a common gateway.

Counting the browsers is pretty trivial... counting those machines which
don't access the net, that's a touch more difficult.  In fact, for an
outsider, it is effectively impossible.

So let's see... you can't tell how many machines are involved... you can't
count downloads (one download could easily be used on all those machines),
you can't count package updates (caching)... how, exactly, do you plan to
determine how many machines are running a given OS, when such a *common*
setup is effectively invisible to your counting methods?

Counting browsers is imperfect, but you can get at least vaguely reliable
stats on it with a little care.  Counting OS installs, by contrast, is a
fool's errand even with Windows, never mind with Linux.

[Date Prev][Date Next][Thread Prev][Thread Next]
Author IndexDate IndexThread Index