<s>
The	O
usage	B-Application
share	I-Application
of	I-Application
web	I-Application
browsers	I-Application
is	O
the	O
portion	O
,	O
often	O
expressed	O
as	O
a	O
percentage	O
,	O
of	O
visitors	O
to	O
a	O
group	O
of	O
web	O
sites	O
that	O
use	O
a	O
particular	O
web	B-Application
browser	I-Application
.	O
</s>
<s>
Measuring	O
browser	B-Application
usage	I-Application
in	O
the	O
number	O
of	O
requests	O
(	O
page	O
hits	B-Protocol
)	O
made	O
by	O
each	O
user	B-Protocol
agent	I-Protocol
can	O
be	O
misleading	O
.	O
</s>
<s>
Not	O
all	O
requests	O
are	O
generated	O
by	O
a	O
user	O
,	O
as	O
a	O
user	B-Protocol
agent	I-Protocol
can	O
make	O
requests	O
at	O
regular	O
time	O
intervals	O
without	O
user	O
input	O
.	O
</s>
<s>
Certain	O
anti-virus	O
products	O
fake	O
their	O
user	B-Protocol
agent	I-Protocol
string	O
to	O
appear	O
to	O
be	O
popular	O
browsers	B-Application
.	O
</s>
<s>
This	O
is	O
done	O
to	O
trick	O
attack	O
sites	O
that	O
might	O
display	O
clean	O
content	O
to	O
the	O
scanner	O
,	O
but	O
not	O
to	O
the	O
browser	B-Application
.	O
</s>
<s>
The	O
Register	O
reported	O
in	O
June	O
2008	O
that	O
traffic	O
from	O
AVG	O
Linkscanner	O
,	O
using	O
an	O
IE6	B-Application
user	B-Protocol
agent	I-Protocol
string	O
,	O
outstripped	O
human	O
link	O
clicks	O
by	O
nearly	O
10	O
to	O
1	O
.	O
</s>
<s>
A	O
user	O
who	O
revisits	O
a	O
site	O
shortly	O
after	O
changing	O
or	O
upgrading	O
browsers	B-Application
may	O
be	O
double-counted	O
under	O
some	O
methods	O
;	O
overall	O
numbers	O
at	O
the	O
time	O
of	O
a	O
new	O
version	O
's	O
release	O
may	O
be	O
skewed	O
.	O
</s>
<s>
Occasionally	O
websites	O
are	O
written	O
in	O
such	O
a	O
way	O
that	O
they	O
effectively	O
block	O
certain	O
browsers	B-Application
.	O
</s>
<s>
One	O
common	O
reason	O
for	O
this	O
is	O
that	O
the	O
website	O
has	O
been	O
tested	O
to	O
work	O
with	O
only	O
a	O
limited	O
number	O
of	O
browsers	B-Application
,	O
and	O
so	O
the	O
site	O
owners	O
enforce	O
that	O
only	O
tested	O
browsers	B-Application
are	O
allowed	O
to	O
view	O
the	O
content	O
,	O
while	O
all	O
other	O
browsers	B-Application
are	O
sent	O
a	O
"	O
failure	O
"	O
message	O
,	O
and	O
instruction	O
to	O
use	O
another	O
browser	B-Application
.	O
</s>
<s>
Many	O
of	O
the	O
untested	O
browsers	B-Application
may	O
still	O
be	O
otherwise	O
capable	O
of	O
rendering	O
the	O
content	O
.	O
</s>
<s>
Sophisticated	O
users	O
who	O
are	O
aware	O
of	O
this	O
may	O
then	O
"	O
spoof	O
"	O
the	O
user	B-Protocol
agent	I-Protocol
string	O
in	O
order	O
to	O
gain	O
access	O
to	O
the	O
site	O
.	O
</s>
<s>
Firefox	B-Application
,	O
Chrome	B-Application
,	O
Safari	B-Application
,	O
and	O
Opera	B-Application
will	O
,	O
under	O
some	O
circumstances	O
,	O
fetch	O
resources	O
before	O
they	O
need	O
to	O
render	O
them	O
,	O
so	O
that	O
the	O
resources	O
can	O
be	O
used	O
faster	O
if	O
they	O
are	O
needed	O
.	O
</s>
<s>
This	O
technique	O
,	O
prerendering	O
or	O
pre-loading	O
,	O
may	O
inflate	O
the	O
statistics	O
for	O
the	O
browsers	B-Application
using	O
it	O
because	O
of	O
pre-loading	O
of	O
resources	O
which	O
are	O
not	O
used	O
in	O
the	O
end	O
.	O
</s>
<s>
Firefox	B-Application
1.5	O
(	O
and	O
other	O
Gecko-based	O
browsers	B-Application
)	O
and	O
later	O
versions	O
use	O
fast	O
Document	B-General_Concept
Object	I-General_Concept
Model	I-General_Concept
(	O
DOM	B-General_Concept
)	O
caching	O
.	O
</s>
<s>
JavaScript	B-Language
is	O
executed	O
on	O
page	O
load	O
only	O
from	O
net	O
or	O
disk	O
cache	O
,	O
but	O
not	O
if	O
it	O
is	O
loaded	O
from	O
DOM	B-General_Concept
cache	O
.	O
</s>
<s>
This	O
can	O
affect	O
JavaScript-based	O
tracking	O
of	O
browser	B-Application
statistics	I-Application
.	O
</s>
<s>
While	O
most	O
browsers	B-Application
generate	O
additional	O
page	O
hits	B-Protocol
by	O
refreshing	O
web	O
pages	O
when	O
the	O
user	O
navigates	O
back	O
through	O
page	O
history	O
,	O
some	O
browsers	B-Application
(	O
such	O
as	O
Opera	B-Application
)	O
reuse	O
cached	O
content	O
without	O
resending	O
requests	O
to	O
the	O
server	O
.	O
</s>
<s>
Generally	O
,	O
the	O
more	O
faithfully	O
a	O
browser	B-Application
implements	O
HTTP	O
's	O
cache	O
specifications	O
,	O
the	O
more	O
it	O
will	O
be	O
under-reported	O
relative	O
to	O
browsers	B-Application
that	O
implement	O
those	O
specifications	O
poorly	O
.	O
</s>
<s>
Browser	B-Application
users	O
may	O
run	O
site	O
,	O
cookie	O
and	O
JavaScript	B-Language
blockers	O
which	O
cause	O
those	O
users	O
to	O
be	O
under-counted	O
.	O
</s>
<s>
For	O
example	O
,	O
common	O
AdBlock	O
blocklists	O
such	O
as	O
EasyBlock	O
include	O
sites	O
such	O
as	O
StatCounter	O
in	O
their	O
privacy	O
lists	O
,	O
and	O
NoScript	B-Operating_System
blocks	O
all	O
JavaScript	B-Language
by	O
default	O
.	O
</s>
<s>
The	O
Firefox	B-Application
Add-ons	O
website	O
reports	O
15.0	O
million	O
users	O
of	O
AdBlock	O
variants	O
and	O
2.2	O
million	O
users	O
of	O
NoScript	B-Operating_System
.	O
</s>
<s>
Squid	B-Protocol
)	O
may	O
have	O
repeat	O
requests	O
for	O
certain	O
pages	O
served	O
to	O
the	O
browser	B-Application
from	O
the	O
cache	O
,	O
rather	O
than	O
retrieving	O
it	O
again	O
via	O
the	O
Internet	O
.	O
</s>
<s>
Websites	O
often	O
include	O
code	O
to	O
detect	O
browser	B-Application
version	O
to	O
adjust	O
the	O
page	O
design	O
sent	O
according	O
to	O
the	O
user	B-Protocol
agent	I-Protocol
string	O
received	O
.	O
</s>
<s>
This	O
may	O
mean	O
that	O
less	O
popular	O
browsers	B-Application
are	O
not	O
sent	O
complex	O
content	O
(	O
even	O
though	O
they	O
might	O
be	O
able	O
to	O
deal	O
with	O
it	O
correctly	O
)	O
or	O
,	O
in	O
extreme	O
cases	O
,	O
refused	O
all	O
content	O
.	O
</s>
<s>
Thus	O
,	O
various	O
browsers	B-Application
have	O
a	O
feature	O
to	O
cloak	O
or	O
spoof	O
their	O
identification	O
to	O
force	O
certain	O
server-side	O
content	O
.	O
</s>
<s>
Default	O
user	B-Protocol
agent	I-Protocol
strings	I-Protocol
of	O
most	O
browsers	B-Application
have	O
pieces	O
of	O
strings	O
from	O
one	O
or	O
more	O
other	O
browsers	B-Application
,	O
so	O
that	O
if	O
the	O
browser	B-Application
is	O
unknown	O
to	O
a	O
website	O
,	O
it	O
can	O
be	O
identified	O
as	O
one	O
of	O
those	O
.	O
</s>
<s>
For	O
example	O
,	O
Safari	B-Application
has	O
not	O
only	O
"	O
Mozilla/5.0	O
"	O
,	O
but	O
also	O
"	O
KHTML	B-Language
"	O
(	O
from	O
which	O
Safari	B-Application
's	O
WebKit	B-Language
was	O
forked	O
)	O
and	O
"	O
Gecko	B-Language
"	O
(	O
the	O
engine	O
of	O
Firefox	B-Application
)	O
.	O
</s>
<s>
Some	O
Linux	B-Application
browsers	B-Application
such	O
as	O
GNOME	B-Language
Web	I-Language
identify	O
themselves	O
as	O
Safari	B-Application
in	O
order	O
to	O
aid	O
compatibility	O
.	O
</s>
<s>
Net	O
Applications	O
,	O
in	O
their	O
report	O
,	O
uses	O
unique	O
visitors	O
to	O
measure	O
web	B-Application
usage	I-Application
.	O
</s>
<s>
The	O
effect	O
is	O
that	O
users	O
visiting	O
a	O
site	O
ten	O
times	O
will	O
only	O
be	O
counted	O
once	O
by	O
these	O
sources	O
,	O
while	O
they	O
are	O
counted	O
ten	O
times	O
by	O
statistics	O
companies	O
that	O
measure	O
page	O
hits	B-Protocol
.	O
</s>
<s>
Net	O
Applications	O
uses	O
country-level	O
weighting	B-Algorithm
as	O
well	O
.	O
</s>
<s>
The	O
goal	O
of	O
weighting	B-Algorithm
countries	O
based	O
on	O
their	O
usage	O
is	O
to	O
mitigate	O
selection	O
area	O
based	O
sampling	O
bias	O
.	O
</s>
<s>
This	O
bias	O
is	O
caused	O
by	O
the	O
differences	O
in	O
the	O
percentage	O
of	O
tracked	O
hits	B-Protocol
in	O
the	O
sample	O
,	O
and	O
the	O
percentage	O
of	O
global	O
usage	O
tracked	O
by	O
third	O
party	O
sources	O
.	O
</s>
<s>
The	O
following	O
tables	O
summarize	O
the	O
usage	O
share	O
of	O
all	O
browsers	B-Application
for	O
the	O
indicated	O
months	O
.	O
</s>
<s>
According	O
to	O
StatCounter	O
web	O
use	O
statistics	O
(	O
a	O
proxy	O
for	O
all	O
use	O
)	O
,	O
in	O
the	O
week	O
from	O
7	O
–	O
13	O
November	O
2016	O
,	O
"	O
mobile	B-Application
"	O
(	O
meaning	O
smartphones	B-Application
)	O
alone	O
(	O
without	O
tablets	O
)	O
overtook	O
desktop	O
for	O
the	O
first	O
time	O
and	O
by	O
the	O
end	O
of	O
the	O
year	O
smartphones	B-Application
were	O
in	O
the	O
majority	O
.	O
</s>
<s>
The	O
two	O
biggest	O
continents	O
,	O
Asia	O
and	O
Africa	O
,	O
have	O
been	O
mobile-majority	O
for	O
a	O
while	O
,	O
and	O
Australia	O
is	O
by	O
now	O
desktop-minority	O
too	O
.	O
</s>
<s>
A	O
few	O
countries	O
in	O
Europe	O
and	O
South	O
America	O
have	O
also	O
followed	O
this	O
trend	O
of	O
being	O
mobile-majority	O
.	O
</s>
<s>
In	O
March	O
2015	O
,	O
for	O
the	O
first	O
time	O
in	O
the	O
US	O
the	O
number	O
of	O
mobile-only	O
adult	O
internet	O
users	O
exceeded	O
the	O
number	O
of	O
desktop-only	O
internet	O
users	O
with	O
11.6	O
%	O
of	O
the	O
digital	O
population	O
only	O
using	O
mobile	B-Application
compared	O
to	O
10.6	O
%	O
only	O
using	O
desktop	O
;	O
this	O
also	O
means	O
the	O
majority	O
,	O
78%	O
,	O
use	O
both	O
desktop	O
and	O
mobile	B-Application
to	O
access	O
the	O
internet	O
.	O
</s>
<s>
StatCounter	O
statistics	O
are	O
directly	O
derived	O
from	O
hits	B-Protocol
(	O
not	O
unique	O
visitors	O
)	O
from	O
3	O
million	O
sites	O
using	O
StatCounter	O
totaling	O
more	O
than	O
15	O
billion	O
hits	B-Protocol
per	O
month	O
.	O
</s>
<s>
Wikimedia	O
traffic	O
analysis	O
reports	O
are	O
based	O
on	O
server	O
logs	O
of	O
about	O
4	O
billion	O
page	O
requests	O
per	O
month	O
,	O
based	O
on	O
the	O
user	B-Protocol
agent	I-Protocol
information	O
that	O
accompanied	O
the	O
requests	O
.	O
</s>
<s>
Note	O
:	O
Wikimedia	O
has	O
recently	O
had	O
a	O
large	O
percentage	O
of	O
unrecognised	O
browsers	B-Application
,	O
previously	O
counted	O
as	O
Firefox	B-Application
,	O
that	O
are	O
now	O
assumed	O
to	O
be	O
Internet	B-Application
Explorer	I-Application
11	I-Application
fixed	O
in	O
the	O
February	O
2014	O
and	O
later	O
numbers	O
.	O
</s>
<s>
And	O
February	O
2014	O
numbers	O
include	O
mobile	B-Application
for	O
Internet	B-Application
Explorer	I-Application
and	O
Firefox	B-Application
(	O
not	O
included	O
in	O
Android	B-Application
)	O
.	O
</s>
<s>
Chrome	B-Application
did	O
not	O
include	O
the	O
mobile	B-Application
numbers	O
at	O
that	O
time	O
while	O
Android	B-Application
does	O
since	O
there	O
was	O
an	O
"	O
Android	B-Application
browser	B-Application
"	O
that	O
was	O
the	O
default	O
browser	B-Application
at	O
that	O
time	O
.	O
</s>
<s>
TheCounter.com	O
is	O
a	O
defunct	O
a	O
web	B-Application
counter	I-Application
service	O
,	O
and	O
identifies	O
sixteen	O
versions	O
of	O
six	O
browsers	B-Application
(	O
Internet	B-Application
Explorer	I-Application
,	O
Firefox	B-Application
,	O
Safari	B-Application
,	O
Opera	B-Application
,	O
Netscape	B-Protocol
,	O
and	O
Konqueror	O
)	O
.	O
</s>
<s>
Other	O
browsers	B-Application
are	O
categorised	O
as	O
either	O
"	O
Netscape	B-Protocol
compatible	O
"	O
(	O
including	O
Google	B-Application
Chrome	I-Application
,	O
which	O
may	O
also	O
be	O
categorized	O
as	O
"	O
Safari	B-Application
"	O
because	O
of	O
its	O
"	O
Webkit	B-Language
"	O
subtag	O
)	O
or	O
"	O
unknown	O
"	O
.	O
</s>
<s>
Internet	B-Application
Explorer	I-Application
8	O
is	O
identified	O
as	O
Internet	B-Application
Explorer	I-Application
7	O
.	O
</s>
<s>
Monthly	O
data	O
includes	O
all	O
hits	B-Protocol
from	O
2008-02-01	O
until	O
the	O
end	O
of	O
the	O
month	O
concerned	O
.	O
</s>
<s>
More	O
than	O
the	O
exact	O
browser	B-Application
type	O
,	O
this	O
data	O
identifies	O
the	O
underlying	O
rendering	O
engine	O
used	O
by	O
various	O
browsers	B-Application
,	O
and	O
the	O
table	O
below	O
aggregates	O
them	O
in	O
the	O
same	O
column	O
.	O
</s>
