tags:

views:

751

answers:

21

Keep in mind that I'm not looking for a list of current browsers to support, I'm looking for logical ways to make that list, backed by some kind of hard statistics.

Since it's been a while since my last web job, I decided to do this latest site up from scratch. Now I have to decide again what to support in terms of browsers. Certainly I have a list of what I'd like to support, but the decisions that went into that list seem to be a little arbitrary to me. Where can I go to get a reliable picture of browser usage and what seems to be a good point at which to cut off an old version of a browser from support?

A: 

My opinion (has always been) build it to the standards and leave it to the browsers to render it correctly.

Unkwntech
A naive approach that I use for personal sites, but does nothing for a commercial venture.
Dustman
Funny, because I own a commercial venture that builds websites for other commercial ventures and it has yet to let me down. If my code is good why the hell should I go through more effort because someone else's isn't?
Unkwntech
"...along the lines of 2-12% based on past experience..." This is somewhat helpful, and I *definitely* do appreciate your experience, even though it's anecdotal."...that is an acceptable loss." Now how would you know that, without knowing what business I'm in? That's not helpful.
Dustman
The problem is that because there is no reference implementation of how a browser should render, there is no true standard. Have you ever read the standards? They're open to interpretation in many places which means no two implementations, even 'standards compliant' ones, will render quite the same.
Greg Beech
@Dustman, I didn't think that it mattered what business you are in because I am averaging from all the data my business has collected, and I am in web design, we have a wide spread client base that spans almost everything you could imagine, and probably some you couldn't
Unkwntech
This isn't really an answer to the original question.
Matt Kantor
@Greg Beach - yes, there IS a reference implementation of a web browser, Amaya from the W3: http://www.w3.org/Amaya/ - it has been around forever, they keep updating it, and it has good support for the standards.
JohnnySoftware
+2  A: 

The browser is dead when (a) a very small percentage of people use it and (b) you don't care about (selling to? educating? whatever your business is) such a small percentage of people.

Kyralessa
By your definition, Chrome is dead...
Daok
It doesn't much matter whether Chrome is considered "dead" or irrelevant. It uses WebKit which is a highly standards compliant renderer. As long as people design their websites to the standard then it will work on Chrome, Safari, Firefox, and other browsers which care about standards compliance.
Wedge
People who use Chrome are probably more tech-savvy than average, and your business may care about selling to such people. Which is covered in point (b).
Kyralessa
A: 

Start with the browser with the highest market share and work your way down from there.

If you have existing metrics on browsers that visit your site, use those instead of the general market share.

Geoffrey Chetwood
How do I find out who has the greatest market share? Good step 1, but what about step 2?Note that I have no current site to get metrics from yet, and see other comments on self-analysis.
Dustman
Google Analytics (http://www.google.com/analytics/) is a great one for metrics on your site. IE6/7 has the highest market share in general if you have no metrics to go off of.
Geoffrey Chetwood
@Rich B: But designing your site in IE first is often the worst approach.
EFraim
A: 

I agree with Unkwntech.

You should try to make the website compatible to both IE and Firefox

David
Coding to the standard (as Unkwntech suggests) doesn't guarantee compatibility with either browser.
Chris Upchurch
Unless you (wisely) consider IE the standard and you don't listen to the standards trolls.
Geoffrey Chetwood
How can IE be the standard when every version renders differently?
Chris Upchurch
Standards change. The W3 'standards' change too.
Geoffrey Chetwood
So why should we code to IE then?
Chris Upchurch
It has the most market share. More market share == more audience. You alienate less people developing for IE.
Geoffrey Chetwood
Total marketshare <> marketshare in your intended audience. For instance, ArsTechnica gets about twice as many visits from Firefox users as it does from IE users.http://arstechnica.com/news.ars/post/20080910-aweek-of-chrome-googles-browser-gets-7-share-at-ars.html
Chris Upchurch
So unless you know what browsers your target audience is using, you can't just assume IE is biggest. For ArsTechnica (or stackoverflow, for that matter) to assume that everyone uses IE would be disastrous for those sites.
Chris Upchurch
Again, as I stated in my post, your traffic (and mileage) may vary. For the vast majority of sites IE is the top.
Geoffrey Chetwood
Even if you assume a more typical distribution, 25% of site visitors are using non-IE browsers. Can any site afford to arbitrarily reduce it's traffic by 25%?
Chris Upchurch
I have not argued you should support any other browsers. That is your assumption.You should almost always target IE first though and support others as needed.
Geoffrey Chetwood
*should not support. That is what I get for not sitting when typing.
Geoffrey Chetwood
But, your not reducing the traffic by 25% if your shit is to the standards, then you will loose more along the lines of 2-12% based on past experience, and that is an acceptable loss.
Unkwntech
Most developers I have talked to (yes, this is just anecdotal evidence) find it easier to develop for non-IE browsers and then apply fixes for IE then the other way around. So supporting IE first creates more work.
David Dorward
Developing for IE is a pain. It has no development tools. You develop in Firefox, and then check in other browsers, applying fixes if needed.
TRiG
A: 

My (very poor) solution was to get stats from w3schools and base my decisions on that. While those numbers aren't really terrible, they are skewed because viewers of that site are more likely to be upgrade-conscious. Also, it doesn't give a breakdown of any browser versions except FF.

Dustman
w3school is pretty bad because it contain stats from mostly developper so FF is too much high
Daok
Absolutely true, but I'm not worried about market share amongst current browsers, just which ones *are* current. And I'd like to have some basis for that opinion other than anecdotal evidence.
Dustman
+1  A: 

See also this topic

Kevin Conner
A: 

If you purely build to standards, some browser won't render correctly since no browser supports all standards. You have to pick a few browsers and test your site in those.

Don't try to be too bleeding edge. If you must use some cutting edge CSS, then you have to expect it not to work 100% of the time.

DoniG
A: 

Yes Unkwntech, you're right with that, but finally is the customer who has to say almost everything about the browser, so if my customer says, for example, Internet Explorer, I will work thinking about IE on my design.

Other thing is if you are making a site for your own business... There you have to check what Dustman requires.

Of course, this is my opinion, and I'm totally agree about doing things under the web standards, for sure.

This should be a comment, not and answer since you are not answering anything and only discussing his post.
Geoffrey Chetwood
+19  A: 

Browsers don't die out completely for about a decade. The first thing you must realise is that you will have some visitors that are using a browser you don't support. The question is not which browsers are not dead, but which browsers are worth supporting (the benefit) relative to the work it takes to do so (the cost).

I've never seen browser statistics I'm comfortable recommending, they all seem to be snake oil. A rule of thumb I feel is appropriate is that a browser isn't worth supporting if somebody using that browser is going to regularly run into problems on other websites as well. In other words "stick with what everybody else is supporting". To that end, Yahoo's graded browser support is useful.

Ultimately, the best choice depends on your individual circumstances and will change over time. For instance, 37signals have recently dropped support for Internet Explorer 6 and Facebook are slowly heading in the same direction. This isn't a decision that most organisations can make yet, but give it a year or two and you'll see a lot more organisations follow suit. Right now, it's a bold step that you probably can't justify, but give it time.

Don't fall into the trap of thinking that supporting as many browsers as possible is automatically the best choice - it may be that you are doing your visitors a disservice by wasting time working on compatibility with a browser used by five people when you could be improving the experience for the other million users you have.

Also, it's worth considering that you can "officially" not support a browser. For example, one thing I've done in the past is use JavaScript served only to Internet Explorer 5.5 and below (via a conditional comment), to automatically remove stylesheets, JavaScript and replace images with their alt text. Without those measures, the site would be unreadable due to Internet Explorer's many layout bugs, but with it, the site at least works, even if it's too much work to "support" it.

Jim
Good browsers never die. They just fade away.
Geoffrey Chetwood
Bad browsers, on the other hand, tend to stick around like the mysterious greenish-brown gunk at the back of the fridge.
Piskvor
+3  A: 

You could take a look at the way Yahoo! supports browsers at Graded browser support.

Peter Stuifzand
+5  A: 

The easiest way to do it is sign up for Google Analytics and add their tracking code to your site (there are a number of similar services, but Google's one is the best I've found). It gives you detailed statistics as to what browsers people who visit your site use.

Once you have a couple of months data, you can start making decisions as to which browsers you will support. I work for a mainstream web company who want to make our site work for as many users as possible, so we consider any browser with above 0.5% market share to be within our testing matrix. However, other sites may choose to only support and test on major browsers such as IE and Firefox.

As a rough guide, the major browsers you'll see are IE 6 and 7, and Firefox 2 and 3. This should cover well over 90% of your audience so is a good starting point for the first couple of months. Then use your analytics data and make a business decision as to whether the potential revenue (or whatever you're trying to achieve) is worth the additional effort it will take to support other browsers.

Added 2008-09-18:

Admittedly one issue with this method is that if your support for some browser types is so bad that your site is unusable with them then it will potentially skew the statistics as those people will stop coming back, and thus those browsers will appear to have a lower percentage of users.

To determine whether this is happening, you can use Google Analytics' detailed breakdown of behaviour for each browser type and version. This gives you the bounce rate, average time on site, pages per visit, and percent of new visits. If the figures for a given browser type and version are significantly worse than others (i.e. the bounce rate is higher, time on site is lower, pages per visit is lower, or percent of new visits is higher) then it's possible that your site isn't supporting that browser sufficiently well and that you might get more users with it if you had better support.

At this point the figures will still give you a reasonable feeling for how important the browser is (i.e. if it you don't support Google Chrome and it is being shown as 2% of your traffic, then it wouldn't jump to 20% just because you added support) so you can use that browser to see how bad your site is, and make a judgment call as to whether you add support; sometimes this may involve fixing only the worst issues and leaving the site imperfect but usable until the browser gets to a higher percentage of users, or out of beta status.

Greg Beech
Kind of overkill to sign up for Google analytics just to get browser brand/version? Could just write a script to pull that information out of the HTTP server logs, and count how many visits from each browser, based on the user agent header value. He will not be able to use any metric collection technique till he has his site up, though. For now, he will have to go with canned metrics from elsewhere, and knowing his target demographic.
JohnnySoftware
A: 

Whichever has < 1% market share.

Imran
+1  A: 

You should use a good UI framework that solves most of the compatibility issues among browsers, like YUI!, jQuery, and so on...

Personaly, I recommend YUI!

Daniel Silveira
Those are very popular UI libraries for JavaScript but his question was really focussed on what browser to support or not; e.g. does "Mosaic" need to be supported - or not?
JohnnySoftware
A: 

What are you really going to do with the list? Are you planning to block browsers you don't support? What if the user hacks the User-Agent response?

Like others, I would strongly suggest going with something like Yahoo's "Graded Browsers" and, if possible, leveraging YUI or other libraries so you don't have to do it yourself.

David Singer
Oh, not at all, it's just that don't want to spend a single moment of my time thinking about, say, Netscape 2. And I thought that since everyone is going to have their own opinion on exactly what to support, I'd like to know how they arrived at those conclusions.
Dustman
+1  A: 

Try to answer this locally, consider your audience. For example when I was developing my own Blog Engine, my appeal was mostly to .NET developers. I hope it stands to reason what browser I primarily develop for. From that point I consider the market share and try to ensure a "reasonable" support level for all other browsers. For example even .NET developers occasionally use Firefox, maybe even Opera. Safari and Chrome are possibilities too now. So my current level of support ranks in this order:

  • It MUST run perfectly in Internet Explorer 7. All features I intended to build are there
  • It MUST run reasonably in Internet Explorer 6, Firefox 3.0, Opera 9+ and Safari for Windows, not everything has to be flawless, but it can't look downright ugly either

Everything else I don't care about. I just don't have the time and willing effort to support everything.

How do I determine whether or not I want to even consider supporting another browser or continuing supporting one of the above browsers any more? Simply I look at the market share and the statistics of who is hitting my page. If someone is dying, or I just haven't seen them in awhile, then I consider support dropped.

So in short, I would simply make a statement to yourself about the browsers that must run your code perfectly then reasonably and update periodically as the browser world changes. For the first run of your website, just think about your audience, for subsequent updates, your statistics should tell you enough.

tekiegreg
A: 

<1% market share isn't a criteria - esp if the browser is new.

For me, < IE6 is dead, and the HTML monkeys I work with WISH it was dead. < FF2 is dead. Opera is a nice to have. < Safari 2 is dead, tho most are designing for Saf 3 now.

So it's: IE6,7,8 FF 2,3 Saf 3,4 Chrome (which is basicly Saf4)

But depending on your app, and how many people you think you are going to get wih hold machines, you COULD drop IE6, which would make your life so much easier.

Nic Wise
+2  A: 

Unfortunately, you won't find a good answer to this; even if you found some hard statistics on browser versions for visitors to your website, that almost certainly doesn't tell you what you need to know.

What you need to know isn't "what percent of my visitors use Browser X", it's "what percent of my revenue comes from visitors who use Browser X". That one guy visiting your site using an ancient copy of IE might be the managing director of a big company wanting to buy a site license; the 10k visitors you had last month using Firefox 3 might be college students wanting to plagiarize your documentation for an essay.

Really, you need to know your market - not just the raw browser statistics. If you pay the bills by selling stuff to graphic designers, then rock solid Safari support matters a lot more than if you're in the job of selling Visual Studio plugins. Not helpful, I know!

Cody Hatch
Vacation activities, with a wide range of pricing and an international audience. Doesn't narrow things down much. :)
Dustman
+2  A: 

There are 2 main groups to target. (There are plenty of others though)

Group #1 is browsers that use Webkit (Safari for example), Presto (Opera for example), KHTML (Konqueror for example) or Gecko (Firefox for example). These browsers should all get the same markup, CSS and Javascript code (as they're all in the same group of standard-compliant browsers). Only work around bugs in one of these if you absolutely have to and have the resources to do so. Instead, test in the latest stable versions of each (as you're developing so they can keep each other in check as to what the expected behavior is) and (after checking in the nightlies for the bugs) file bug reports. Again, avoid workarounds for a specific browser if you can. Instead, plan a cross-browser compatible solution from the beginning.

With Group #1, you don't have to worry about older versions much, if it all.

Group #2 is browsers that use Trident (IE for example). Target IE versions you care about and still only workaround the most severe bugs.

Also, don't deny browsers you don't officially support. Let them fend for themselves instead of blocking them (either intentionally or through crappy browser detection).

Also, remember that when looking at market share percentages, try to figure out the numbers they represent so you can see how many millions of potential visitors with that browser there are. 1% or 5% might not seem like a lot, but that could still mean millions.

Most of all, listen to the visitors. If you're getting multiple complaints about a certain browser, look into it if you can. Even if it's for a browser with low market share, if it's a trivial fix, you should just do it.

Ones that are definitely not dead are: IE6 (starting to push it), IE7, IE8, latest Opera 9.x, latest FF 3.x, latest Safari 3.x and others that have about the same capabilities. FF 2.x isn't dead either and is needed for Win9X users (if they don't want to use Opera)

Shadow2531
A: 

It's simple - most users keep using the browser that came with the PC when they bought it (think of your mom). The browser is dead when the machines that it pre-installed with are not longer used for Internet access... which is probably around 5 years. As prices of new PC's drops and they become more of a consumer electronics item then this period will drop as people will easily buy a new PC

Simon Munro
IE6 has lingered a lot longer because Microsoft fused it to the OS and made it impossible to support more than one version of IE on the same computer, at least for XP. People say newer Windows versions treat IE more like an app - but IE6 is like a stain that is taking way more than 5 years to go away. Firefox 2 is rapidly vanishing because 3 and 3.5 are better and 2 was not really "coupled" to an OS. Also, IE6 had serious presentation flaws. Microsoft was the only company to get the W3 standard box model *wrong*. Some trick, considering the standard spelled out how it was to be defined.
JohnnySoftware
A: 

I would say IE6 and below are dead... but many are still stuck using it.

This site has a nice live listing of each browser and its actual age.

http://webbugtrack.blogspot.com/2008/08/browser-life-statuses.html

scunliffe
A: 

I'd go with the http://browser-update.org/ defaults, which currently say the following are dead:

IE <= 6 FF <= 2.0 Op <= 10.01 Sf <= 2.0

TRiG