Where can I get a good list of browser market shares

David Walker wrote:

WRONG: 93% of all hits came from Internet Explorer.
RIGHT: 93% of all recorded hits came from Internet Explorer.

No - that's still wrong. Try:
93% of all recorded hits came from agents identifying themselves
as Internet Explorer.
(And even the correctness of that statement would need to be reviewed in
the light of the exact method being used.)

Thats actually 93% of hits from agents with "MSIE" somewhere in the browser.
If you test for other browser names, they are almost always in the string
too and so you can put these down as the real browser if you test enough.

So, we're measuring the percent of all browsers being used to access a
site, except that we can't, so we're measuring all browsers that can
be measured -- an exercise in circular logic, to be sure -- but
browsers sometimes use a faked ua string. But the real id is "almost
always in the string." Almost? So what percentage of browsers have
the real name in the string? What percentage didn't fake it and what
percentage did?

Meanwhile, this thread has ignored the other factor making browser
market share stats misleading: caching, including proxy caches.

This data is looking less and less reliable all the time.

--
Brian
follow the directions in my address to email me

Shawn K. Quinn wrote:

Which, of course, we really should not have to do. It's ridiculous! What
happened to letting the browser just identify itself as what it really is
rather than some black magic string that pays tribute to other types of
browsers just because?

What happened is clueless coders started producing content based on
which browser was being used. If we had a nickel for every site whose
browser sniffer was broken... Sometimes I think the webmaster is the
only one successfully sniffing, and it ain't Microsoft products up his
nose!

Remember, just because a browser identifies itself as something, doesn't
mean it really is. The Junkbuster proxy and a feature in Konqueror automate
the process of masquerading as other browsers.

I often surf while sending no referrer header at all.

--
Brian
follow the directions in my address to email me

On Mon, Jul 14, Shawn K. Quinn inscribed on the eternal scroll:

It's ridiculous! What
happened to letting the browser just identify itself as what it really is
rather than some black magic string that pays tribute to other types of
browsers just because?

What happened is that some control-freaks decided to over-interpret
what it said in the RFCs (this was RFC2068 - RFC1945 was substantially
the same):

This is for statistical purposes, the tracing of protocol
violations, and automated recognition of user agents for the sake
of tailoring responses to avoid particular user agent limitations.

- and decided instead of merely "avoiding particular user agent
limitations", to go right ahead and comprehensively re-engineer their
content according to what the user agent claimed to be.

This in spite of (or in some cases, sadly, even _because_ of![1] - the
effects of server-side negotiation - already warned about, for
example, in RFC2068 section 12.1.

RFC2616 says much the same, of course...

[1] some of these control freaks think that damaging the web's ability
to cache is the greatest thing since sliced bread...)

Remember, just because a browser identifies itself as something, doesn't
mean it really is. The Junkbuster proxy and a feature in Konqueror automate
the process of masquerading as other browsers.

Ah, but the control freaks will tell you that if you masquerade as X,
then you deserve to be treated as X. If it only meant avoiding some
of the more spectacular browser bugs (as indeed the RFC suggests) then
it might not be so bad - although there are often much nicer ways of
achieving that (I'm thinking of those various CSS-hiding tricks to
persuade browsers to ignore parts that they would find indigestible).

But the number of times I've been told to downgrade my browser to e.g
NN4 in the name of getting a "proper" browser (and have seen the
indexing robots lectured in the same way!) shows at least one of the
unattractive featuers of this control-freak approach.

In article <HmUPa.266$2p4.103 (AT) newssvr19 (DOT) news.prodigy.com>,
David Safar <gwythinn (AT) pacbell (DOT) net> wrote:

What better measure of what software and settings your visitors use
than data collected from your visitors themselves?

Well, the problem with that is that it's not a given you
want to know what your current visitors use, but rather
what your _potential_ visitors use.

It's also pretty meaningless unless you've got a method to
determine what the *first*time* visitor subset uses. No,
it's actually totally meaningless.
--
Urban Fredriksson http://www.canit.se/%7Egriffon/
A king and an elephant were sitting in a bathtub. The king said, "pass
the soap" and the elephant said, "No soap, radio!"

Brian wrote:

Shawn K. Quinn wrote:
[...]
Remember, just because a browser identifies itself as something, doesn't
mean it really is. The Junkbuster proxy and a feature in Konqueror
automate the process of masquerading as other browsers.

I often surf while sending no referrer header at all.

I have as well, from time to time. For a very long spell I also had
Junkbuster identify my User-Agent as "Not Available".

There are a few sites that rely on the optional Referer header, which is
also quite broken behavior that I didn't even touch upon in that post.

--
Shawn K. Quinn