Log in

LiveJournal user agents - brad's life [entries|archive|friends|userinfo]
Brad Fitzpatrick

[ website | bradfitz.com ]
[ userinfo | livejournal userinfo ]
[ archive | journal archive ]

LiveJournal user agents [Mar. 26th, 2005|12:53 pm]
Brad Fitzpatrick
The cool thing about LJ's logging system is that we can attach to it and subscribe to the stream of hits and do realtime stats. Yesterday I wrote a bunch of quick regexps to infer browser versions and operating systems from user agent strings. (doing things like checking for "Safari" before Gecko because Safari advertises that it's "like Gecko")

The results: (20 second sample, removing all LJ clients and web spiders....)

31059 MSIE 6.0
10473 Firefox
1438 Safari
1263 Opera
892 MSIE 5.5
341 MSIE 5.0
208 MSIE 5.01
166 MSIE 5.23
148 MSIE 5.22
124 MSIE 4.01
121 MSIE 5.17

And operating systems: (about 15 seconds)

26130 Windows
1566 Mac
304 Linux

Linux may only have 1%, but at least FireFox is kicking ass up there at like 25%.

[User Picture]From: fanless
2005-03-26 09:00 pm (UTC)
I am using Firefox on my Mac.
I am a minority.
(Reply) (Thread)
[User Picture]From: imgreen
2005-03-26 09:04 pm (UTC)
holy shit, that's only 20 seconds?

ew, I can't even use IE anymore (since getting firefox). I think if most people were forced to use it for a week or so they wouldn't go back. bleh.
(Reply) (Thread)
[User Picture]From: crschmidt
2005-03-26 09:19 pm (UTC)
I'm not sure this has a huge affect on LJ, but Firefox does perform a significant number of requests when a favicon can't be found. On subdomains, for example, you could see a request every time someone switches tabs in firefox, because it tries to fetch the icon and cache it, but it won't cache the 404s.

Since LJ has a favicon on the main site, this probably isn't nearly as big of a deal as it could be, nor am I sure if you're including 404s in your stats or anything, but just something to think about, and something that friends of mine have seen as statistically significant in watching browser stats.
(Reply) (Thread)
[User Picture]From: brentdax
2005-03-26 09:24 pm (UTC)
Wow, you get a lot of hits. But we knew that already...

Despite the fact that LJ has an image as a place full of dumb, angsty highschoolers, I'm often surprised at the technical sophistication of many users. It has a large number of Firefox users (and 494 people listing Firefox as an interest), and I'm constantly running into hackers I know from Perl Monks, perl6-internals, or other important communities.

I think the reality is that LJ is for people who don't know how or don't want to bother to set up their own blog. (Or people who think the friends page is the cleverest thing anybody has ever done on such a service. Which it is.)
(Reply) (Thread)
[User Picture]From: granting
2005-03-26 09:37 pm (UTC)
It is the friends' page.

The friends' page is actually why the LJ code base is being considered for use within certain federal government communities.
(Reply) (Parent) (Thread)
[User Picture]From: garrettpalm
2005-03-27 12:23 am (UTC)
Friend's page is the main reason I stick around here. I'm not that tech savvy, but I love the friends page.
(Reply) (Parent) (Thread)
From: evan
2005-03-26 11:50 pm (UTC)
494 firefox interests / 6576793 users = 0.007% of LJ
(Reply) (Parent) (Thread)
[User Picture]From: brentdax
2005-03-26 11:54 pm (UTC)
Yeah, but casual users aren't likely to list a browser in their interests.
(Reply) (Parent) (Thread)
[User Picture]From: brad
2005-03-27 12:10 am (UTC)
That stat means nothing.

I use toilet paper but I don't list it as an interest. It's just what gets the job done.
(Reply) (Parent) (Thread)
From: evan
2005-03-27 01:45 am (UTC)
"That stat means nothing." was my point, too
(Reply) (Parent) (Thread)
[User Picture]From: d4b
2005-03-27 02:56 am (UTC)
LOL!! What a quotable line!
(Reply) (Parent) (Thread)
[User Picture]From: crackmonkey
2005-03-28 03:42 pm (UTC)

That's almost .01%!
(Reply) (Parent) (Thread)
[User Picture]From: janinedog
2005-03-26 09:24 pm (UTC)
Opera can very easily be made to identify itself as something else (usually as IE, so that websites that force you to use IE will still load in Opera). It's an easily accessible menu option. So it's quite possible that there's more Opera users out there than these stats show.

Though, checking it out, the user-agent apparently ends up looking like "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1) Opera 7.54 [en]". So maybe you're already checking for cases like that. :)
(Reply) (Thread)
[User Picture]From: primeling
2005-03-26 09:33 pm (UTC)
I only use Firefox because it has tabs and I can shut off images.

As for OS, I'm cheering for Mac. Though currently I'm on my PC right now. I've noticed some flaws in LJ with OS X, most of them are caused by bells and whistles and aren't compatiable.

At least LJ isn't like gmail and gives OSX the complete bird! I just don't use gmail just because it refuses to accomodate to Mac users.
(Reply) (Thread)
[User Picture]From: chris
2005-03-26 09:48 pm (UTC)
really? I've never noticed any problems using gmail on OSX.
(Reply) (Parent) (Thread)
[User Picture]From: primeling
2005-03-26 09:51 pm (UTC)
For me it says that if I remember correct, an active X problem. I'm not the only one. I have set up my gmail account to accept pop access, and I use gmail as a back up email.
(Reply) (Parent) (Thread)
[User Picture]From: chris
2005-03-26 10:46 pm (UTC)
huh, i use firefox on OSX and gmail behaves exactly the same as it does on XP. what have you noticed that doesn't work right?
(Reply) (Parent) (Thread)
[User Picture]From: crschmidt
2005-03-26 11:25 pm (UTC)
They're probably problems with Safari, rather than OS X as a whole: It seems that for each Google tool, there's some important part of the Javascript stuff it uses that Safari doesn't support. Gmail is the least so out of the group: It *works* in Safari, just crappily. (If you hit back, it confuses gmail to the point you have to drop to the main URL to get back in again, stuff like that.)

XSLTProc is the one that Google Maps uses which Safari doesn't support. Orkut had some feature which didn't work in Safari. Etc.
(Reply) (Parent) (Thread)
[User Picture]From: j7xz49br3m93xrr
2005-03-26 10:10 pm (UTC)
GMail is fine on the Mac in Safari. So's Google Maps, which is even more advanced. GMail even seemed fine to me back when Google claimed they weren't supporting Safari (a long time ago now). Not sure why you're hitting problems, but rest assured.. it's all cool.
(Reply) (Parent) (Thread)
[User Picture]From: duskwuff
2005-03-26 10:28 pm (UTC)
GMail works just fine under Safari.
(Reply) (Parent) (Thread)
[User Picture]From: bevived
2005-03-26 09:49 pm (UTC)
Yay for Firefox. I guess it is not representative for all internet users, but at least LiveJournal (which is my favorite site on the net), so it's all good.
(Reply) (Thread)
[User Picture]From: remark
2005-03-26 10:38 pm (UTC)
Does it toss Mozilla and Netscape in with FireFox, or did nobody use those browsers in that time period?
(Reply) (Thread)
[User Picture]From: brad
2005-03-26 10:42 pm (UTC)
I lumped all "Gecko" together into Firefox because it was mostly Firefox.
(Reply) (Parent) (Thread)
[User Picture]From: lithiana
2005-03-26 10:45 pm (UTC)
you know what'd be neat? a page displaying these stats in real time. with pretty graphs and stuff.

(btw, konqueror advertises itself as "KHTML; like Gecko" as well, but i guess that's not a significant %age of the total.
(Reply) (Thread)
[User Picture]From: grumpy_sysadmin
2005-03-27 01:37 am (UTC)
While it's interesting, I'm not convinced that a ~20 second sampling is sufficient for a real metric here. I'm not stating this as fact, but I would be unsurprised to learn that this statistic was drastically different between 15:00:00 and 15:00:20 EST than it is between 23:50:00 EST and 23:50:20 EST.

I think you could get statistically valuable metrics by sampling 20 seconds once an hour on or about the hour.
(Reply) (Thread)
From: evan
2005-03-27 01:46 am (UTC)
on the hour is more likely to pick up hourly bots. ;)
(Reply) (Parent) (Thread)
[User Picture]From: jwz
2005-03-27 03:43 am (UTC)
If you only counted logged in users, that would probably exclude almost all bots. Comparing the "logged in" and "not logged in" stats might be interesting.
(Reply) (Parent) (Thread)
From: russssss
2005-03-27 11:16 am (UTC)


On Audioscrobbler, Firefox is a good 2% ahead of IE, and Linux has 4.4%. Our userbase is leeter than yours ;).

How does your logging work now then? We've just changed to using mod_log_sql with mod_proxyfixup, then feeding the database into AWStats hourly, after we do some realtime stuff with it.
(Reply) (Thread)
[User Picture]From: mart
2005-03-27 12:20 pm (UTC)

I'm almost certainly not included in these stats anyway, but my browser appears as "Mozilla/5.0 (Compatible)". What would you class that as, per your regex? ;)

(Reply) (Thread)
[User Picture]From: brad
2005-03-27 06:38 pm (UTC)
You'd be ignored.
(Reply) (Parent) (Thread)
[User Picture]From: avatraxiom
2005-03-27 01:09 pm (UTC)
brad: Linux may only have 1%...

Yeah, but we're definitely the coolest 1%.

And we get to experience the glory of LogJam. So there. Hahahaha.

(Reply) (Thread)
[User Picture]From: youngoat
2005-03-28 05:05 pm (UTC)

Browser Stats

Those stats match up pretty closely to w3schools.com's Browser Statistics, though they show Linux as being around 3%. I wonder if your regexp needs tweaking.
(Reply) (Thread)