Switch default testing to IE8?

Just throwing feelers out to see what people think. Should I switch the default browser when people submit tests over to IE8 or leave it with IE7?

I usually like using IE7 for testing because it exposes problems that the more modern browsers hide (well, optimization opportunities anyway) and it still has a reasonable chunk of the market.

Joshua Bixby also has a blog post on why he uses IE7 for his testing: http://www.webperformancetoday.com/2010/10/13/why-test-with-internet-explorer-7/

Here are the stats from StatCounter: http://gs.statcounter.com/#browser_version-ww-monthly-200910-201010



I’ll add a few thoughts.

It’s definitely important to keep IE7 in mind when optimizing. And if the goal is to get people thinking about the tail of the latency distribution, then yes, IE7 is the right browser to optimize for.

But the statscounter chart shows that IE6/7 populations are shrinking quickly. A year ago 35% of users were on IE6 or 7. Now 19% of users are using these browsers. In a year, if we follow the trends on the statcounter page, we’re looking at well under 10% of users on these browsers. On the other hand the footprint of modern browsers increases every day.

Ideally optimizers would look at waterfalls for both IE7 and IE8. But if you’re only going to think about one of the two, you have to ask yourself: do you care about optimizing for the majority of your users, or do you care about optimizing for the tail of your latency distribution? If the tail is important, focus on IE6 and IE7. But keep in mind that these browsers are effectively going away (look at the trends on statscounter before stating otherwise), so I would discourage fine-tuning sites for these old browsers. On the other hand, modern browser features like data URIs, increased parallelization, speculative JS fetching, are only becoming more common, so investing in them will pay off down the road as more users transition to modern browsers.

I agree with Bryan, it really depends on your market and who you want to optimize for, but doing deep dive fine-tuning for IE6/7 is probably going to show diminishing returns given the speed at which their market share is declining.

When it comes to WebPagetest specifically and the default test browser I’m in the IE7 camp. Someone who is a WebPagetest master will know what they want to optimize for, and they can pick IE7 or IE8 depending on what they are looking for. Someone who is brand new should probably be shown the worst case scenario. Like you said IE7 exposes problems that more modern browsers hide. I feel like it’s good to show these problems to someone who is new to the web performance field.

Thanks for the plug, Patrick. I’ll be keeping my eye on your poll results. :slight_smile:

hi guys,

Reasons to keep IE7 the default:

  • show the people are new to WPO and Webpagetest the worst scenario (valid point by jklein)
  • for many sites that target the average user, IE7 still has a market share of >10%, and I find that significant.

For these two reasons, I voted for keeping IE7 the default.

Having said that, switching to IE8 might be good, because it is the browser that has a far larger than market share than IE7. At some point (perhaps in 6 to 12 months from now), it will make sense to switch, as IE7 market share drops to 8% or so…

Personally (and being an experienced WPO consultant), I use IE8 more than IE7. Usually this means running a bunch of tests in IE8 and just a few in IE7, only to see if any remarkable stuff shows in the waterfall charts.

In some countries like Thailand, IE7 seems to enjoy > 20% marketshare.

Unless im optimizing for such regions, I usually optimize for IE8 while keeping IE7 in mind, not bothering too much about it.

Id vote for making IE8 default…

And China is all IE6 :slight_smile: (or should that be a massive :frowning: ). 57% according to Global Stats Counter. It’s amazing what pirated XP + no updates will do.

Yeps… last i checked(4 or 5 years ago) in Thailand pirated windows xp was sold for 100 Baht ($3.24). Complete with regionalisation, antivirus installed and setup, preset wallpaper of sexy asian models and common locally used softwares . In fact one of the popular pirated mixes had firefox installed as default browser. That let to a sudden increase in Firefox usage. We figured that trying to explain a sudden spike in firefox usage locally…

South Korea is another sad story. Reasoning explained in this excellent blogpost.

I’m still on IE7 camp for WebPageTest defaults as the goal for people using defaults is to get a picture they do not expect and to see the results in their worst.

I do however agree with Bryan that optimization should be done for newer browsers, but until WPT has more metrics that cover those features (e.g. dataURIs, local caches and so on), all we have is network and showing worst case scenario on the network by default is probably better.

Maybe making browser version stick for a logged in user (not for everybody, but only for those people who went through registration step) is good so people who use the tool often don’t have to switch to the right browser all the time.

Maybe it should be also a setting in the account instead of automatic feature.