WebKit on Mobile, take 2

When I reviewed the reactions to my There is no WebKit on Mobile post, it became pretty clear that few had expected its conclusion that there is no single WebKit on Mobile. Overall, it seemed that most people were pretty surprised, and hurried to revise their ideas of the mobile browser market. That was the point of the article, so I was happy.

The most-often heard criticism was that I was unclear about the browser version numbers. That’s true, and I have updated the table to include them. I also split out the tests into Acid, CSS2, CSS3, HTML5, and JavaScript, and calculated separate scores for each browser. The results are interesting for some browsers. Konqueror sucks at JS but is very good in CSS, while Android is exactly the opposite. Interesting data.

(I’m still tinkering with the interface, by the way, and I didn’t have the time to finish my current revision. So the coloured bars are temporarily gone, but they’ll return in the future.)

Full speed

By far the most interesting and important reaction came from Alex Russell, who has the annoying habit of being right. He wrote WebKit, Mobile, and Progress, and while he does not contest my findings, he objects to my interpretation.

Basically, while I said that the WebKit on Mobile glass is half empty, Alex maintains it is half full. Besides, and this is his point, the interesting part is not the fullness of the glass (which is what I measured) but the speed with which it gets fuller.

I agree mostly; in order to get a real feel for what’s happening on mobile we have to be able to measure progress.

So what I want to do is compare mobile browsers that have been released in the same year (or in the same quarter, or whatever). Sadly, right now I don’t have the time to implement a nice data interface that allows me to draw conclusions, so I have to postpone this experiment. Still, this is something that needs to be done.

Rate of development

In addition, Alex feels the mobile browsers are developing at a faster rate than the desktop browsers. This could be true. First of all the mobile browsers have more stuff to catch up with, and in order to make the mobile web truly break through into everybody’s life the browsers have to get a lot better.

Besides, some HTML5 features, notably geolocation, appcache, and the online/offline events, are mildly interesting on the desktop, but absolutely vital on mobile. Mobile browsers therefore have a larger incentive to implement them soon.

Finally, it’s IE that’s keeping the desktop’s average down. I wonder what would happen if we’d take IE out of the equation. Would mobile browsers still evolve at a faster pace than the desktop ones?

Missing numbers

Alex’s most important point is this:

PPKs data is missing some other columns too, namely a rough estimate of the percent of mobile handsets running a particular version, [and] rates of change in that landscape over the past 18 months.

He’s absolutely right. I should have included these numbers. Problem is: I don’t have them. Nobody does.

Well, that’s not entirely true. The AdMob mobile advertisement platform regularly publishess mobile metrics, and I always study them, partly because they’re genuinely interesting, and partly because there is nothing else.

I give huge amounts of kudos to AdMob for publishing these figures. In fact, I’d willingly borrow kudos at a prohibitive interest rate in order to give them to AdMob.

Still, there are a few problems that preclude their analyses from being exactly what I need:

Thus this source of data, which is by far the best I’ve found so far, does not suit our needs. If you know of more good, free data, please leave a comment.

Until we have a good data source, we just do not know the market share of the various WebKits, or of most other mobile browsers, in fact. We can only be certain of Safari, whose market share coincides exactly with the iPhone’s, but even there we need version numbers.

Turnover rate

Alex says that the turnover rate of mobile devices is somewhere in the order of 24 months. From anecdotal evidence I gathered (mainly among my mobile-web-illiterate friends) it seems that this rate may be significantly longer. Without good data, however, we cannot tell who of us is right.

Still, turnover rate is very important. Right now I spend a lot of time on S60v3, which was created back in 2005 (Alex’s date), and which was an insanely popular OS for quite a few years. I’m not sure when Nokia stopped producing S60v3 devices in favour of the S60v5 touchscreens, but it wasn’t long ago. (If you know a date, please leave a comment.)

From what I gather the S60v3 WebKit is still pretty important (provided its owners surf the Internet, which is not certain). This clashes partly with Alex’s turnover rate, and shows once more how vitally important it is to get good data on mobile browser use. Is S60v3 still very important, or is it being rapidly superseded by newer Nokias? I just don’t know.


Several people asked me why I didn’t include Safari on Mac in the table. Simple: I don’t own a Mac, and it seems unlikely I ever will in the future.

Somewhere around Safari 1.3 I gave up on Mac. I was fundamentally unwilling to pay a lot of money for an OS upgrade every time a new Safari was released, because I used my Mac only for testing purposes, and not for any other reason. Besides, Apple is the only browser vendor that requires you to pay for a new browser (plus other stuff, admittedly, but still). I’m sick of that behaviour. #appleisevil

As a result I wholly skipped Safari 2, and only restarted my Safari testing when the Windows version was released. Nowadays I only test Safari on Windows.

I have never yet heard of a difference between Safari on Mac and on Windows, and besides, the whole point of the Windows release was allowing non-Mac web developers easy access to the browser. Apple would seriously damage itself if there were any differences between the Mac and Windows versions.

Somebody else asked me why I haven’t included iCab. Although the answer is again that I don’t own a Mac, I must admit I made a mistake long ago. When iCab 5 (right?) was released and it turned out to use WebKit as a rendering engine, I stopped testing it because I assumed it would be equal to Safari. In hindsight, I understand I might have found some interesting differences between the two. Too late now.

Advanced Android chaos

Finally, a not-wholly-related but not-wholly-unrelated-either item. Android in particular is going to become a chaotic OS browser-wise.

Yesterday I briefly tested the HTC Hero, which ostensibly runs Android 1.5 . I found that:

The other Androids I tested (which were, incidentally, also HTCs) did not have these features.

If Vodafone has an HTC Hero I plan to do some tests next week, but I can already tell you HTC Android WebKit 1.5 is not necessarily the same as HTC Android WebKit 1.5. Isn’t web development fun?

This is the blog of Peter-Paul Koch, mobile platform strategist, consultant, and trainer. You can also follow him on Twitter.
Atom RSS

I’m speaking at the following conferences:

(Data from Lanyrd)




Comments are closed.

1 Posted by Ed on 13 October 2009 | Permalink

The last few versions of Safari for the Mac have been released for older versions of OS X too. At least versions 3 and 4 have.

Microsoft do the same thing - you can't get IE 8 for Windows 98, ME or 2000, nor could you get IE 7 for these...

2 Posted by Cliff on 13 October 2009 | Permalink

Hi Peter-Paul,

Sadly there is a difference between windows safari and mac safari. This we have experienced with a customer who used a mac...


3 Posted by Sander Aarts on 13 October 2009 | Permalink

I believe it was iCab 4 that switched to WebKit.

Too bad in my opinion, cause it was always fun to see a browser that was deleveloped by only one person and which had better support for some web standards that IE ;)

4 Posted by Arleen Bothwell on 13 October 2009 | Permalink

I'd love to see IE catch up, or humbly fade away.

That's remarkable, 24 months. Not the fault of the o/s or browsers, though, but the onward push of marketing and competition (god bless 'em both).

5 Posted by Joe Lewis on 13 October 2009 | Permalink

An additional thought: It is likely a good portion of mobile users won't update their OS or browser until they purchase a new device. Whatever browser version that shipped with the device is likely where many of them will stay. Great variance in the implementations of WebKit and other rendering engines on mobile platforms is likely to remain a wildly diverse arena.

6 Posted by Andrea Giammarchi on 14 October 2009 | Permalink

if I am not wrong, there is still no -khtml- test but only -webkit- quite unfair for Konqueror. Anyway, I wonder if you can create a sort of test page wrapping info via JavaScript, OS and UA, letting people test them. That would be a good platform for any kind of table related test.
If you need any help with that just let me know. Thanks again for the good work.

7 Posted by Tim Tepaße on 14 October 2009 | Permalink

CSS 3 3D Animations are Webkit/OSX-only for the moment.

8 Posted by Wayne Pan on 14 October 2009 | Permalink

Hey PPK, I work for AdMob. Let me know if you need more granular data. The metrics report is watered down for a general audience.

I'd be happy to get you some better data. Send me an email, it should be attached to this comment.

9 Posted by Speednet on 14 October 2009 | Permalink

RE: Safari Win vs. Apple

The only major difference that could possibly warrant special testing on a Mac is font support. I'm not sure if that would impact any of your tests, but it is a fundamental difference between Mac and PC.

Many sites that are never tested on a Mac end up looking much different because the developer did not use a list of CSS font names with the closest matching Mac names. For example, a common CSS font list is Arial, Helvetica, sans-serif.

10 Posted by Torgeir Lebesbye on 15 October 2009 | Permalink

As a mobile enthusiast, I can't help to notice some minor errors about the mobile OS landscape: Nokia S40 is not built around Symbian, and I would not say that S60v5 is the new S60v3; it's more like yet another feature pack, adding touch screen capability. At the moment, there is about three or four S60v5 handsets available while S60v3 handsets keeps on coming (e.g. E55, N86 and 6760). The next S60 will be Symbian^2 by Symbian Foundation.

And while you're at listing Symbian UIs, DoCoMo's MOAP(S) should be mentioned due to its large sales volume. I'm not sure how it looks today, but it used to be shipped with Opera Mobile a few years back.

11 Posted by Steven Simmons on 16 October 2009 | Permalink

"Besides, Apple is the only browser vendor that requires you to pay for a new browser".

If by that you mean they don't support older operating systems with new browser releases, I think that is a HUGE leap.

IE6 dropped support for Windows 95
IE6.2 dropped support for Windows ME, 98, and NT 4.0
IE7 dropped support for Windows 2000

Firefox 3 dropped support for Windows 98 / ME, as well as OS X 10.3 and lower - which was still in widespread use at the time.

So, even Mozilla engages in this so-called evil behavior.

I think the big regret is you can't run older or multiple versions of the same Safari on an OS - its too integrated with the system, much like IE is on Windows.

12 Posted by Daniel Parks on 18 November 2009 | Permalink

Apologies for the late comment.

You can in fact run multiple copies of Safari on the same system: http://michelf.com/projects/multi-safari/

I don’t know if you can do something similar on Windows.