Test, test, test

Home > Build > Design > Articles > Page View

by Aaron West

If only putting up a web site were as easy as throwing together a few images, typing a few words, and uploading a few files. There is oh so much more. The greatest obstacle to a web designer is cross-platform compatibility. There are plenty of tricks to make pages viewable by other browsers or varying screen resolutions, such as percentage table widths and redirectors. Are these fool proof? No. Different platforms and different browsers handle and see everything a little differently. The only way to guarantee a tight layout is to roll up your sleeves and test it yourself again and again.

At our company, we've put together a strict testing regimen that we feel comfortable with. As a result, (as far as we know) all our pages are viewable the way we want them to be by 99% of all browsers. This might cause more work than necessary, but at least we know our pages will look good regardless of where they're viewed. We think that makes it worth it.

Testing Resolutions, Browsers and Platforms

The first test should always be different resolutions. Most users today use 800 x 600, so it's best to optimize for that resolution. I use a 14 inch monitor for designing only for that reason. It's easier for me to optimize for my target audience if I'm the same audience. That doesn't mean that pages should be optimized only for that resolution, but just that's where they should look the best. Change resolution to 640 x 480. This resolution is where design most often comes apart. Many developers feel that a horizontal scroll bar is a sign of poor, sloppy design. Others feel that people at resolution that low are prepared to scroll a little anyway. It's your design, and your call. All that really matters is that the design stays intact.

Once all resolutions have been tested we must move on to different browsers. I recommend at least testing on the majority of MSIE and Netscape products. MSIE 4 and 5 are now the two most popular browsers, with Netscape 4.5 lagging shortly behind. All of these browsers must be tested at every resolution.

We go further than that and recommend that you do as well. We test on earlier versions of the popular browsers, as well as other browsers that are used. Below is our browser test list, with links to the download page.

Microsoft Internet Explorer: MSIE is now the most widely used and probably the strongest browser available.

Netscape: Netscape still has a great deal of popularity, but it has been steadily decreasing in user base. This may change upon the release of Netscape 5.

Mozilla: This is the open source project for Gecko, Netscape 5's browser engine. It's not a bad idea to download the latest build and testing your site with it.

Opera: Opera is a small and fast browser. Opera is geared toward the professional surfer. It even touts itself as a browser not for entertainment, but for work.

Icab: This small browser is offered only for the Mac, and is quietly gaining popularity. In addition to being a good browser, Icab also strictly checks HTML tags on sites.

I also recommend that you test your sites on a Mac. The Mac versions of MSIE and Netscape render pages differently. I've tested many pages that looked fine on the PC, but came apart on the Mac. Here is our list of platforms to test on:

Windows 95/98: Take your pick.

Windows NT: We've found that most pages under NT look the same under 95/98, but if you have access to an NT machine, it's worthwhile to test there.

Macintosh: The Mac only has 2.5% market share, but is climbing due to the I-Mac's popularity. This number can change due to the content of the site. For example, this site gets quite a few Mac visitors, since there are many Mac web developers out there.

Linux: Linux supposedly composes less than 1% of the market. I don't trust that stat. Linux has shown up in my access logs quite frequently. If you have access to a Linux box, test. If not, then it probably isn't worth it.

Decisions will have to be made

There is no way to please every potential user without compromising something. Sometimes it could be only 5 pixels on an image, whereas other times it could be a nested table. The decision on whether isolating a percentage of users is worth keeping the design you like.

Recently we did a site where we ran into a problem with Macintosh running Netscape 4.6 at 1280 x 1024. Our design came apart when viewing the page at full screen. The odds aren't too high that we would receive many hits with this configuration, so we weren't sure at first whether we should 'tweak' the site to accommodate it. What was most important to us was whether changing the site would affect any higher percentage users. It did, but the change was hardly noticeable. We then estimated how much time it would take to change the rest of the site. We predicted less than 30 minutes, so we made the change. Many people would not have made the change, and that would be fine. We estimated that the likeliness of this combination hitting the site was 1 in a 100,000. We had already estimated that the site would likely get between 20,000-50,000 hits a month. That meant that once every 2 months, the design would fall apart on somebody.

We made that change because we want our design to stay tight at every combination imaginable. Now we can say that the layout is as tight as we could have possibly made it. Other developers might feel differently, and that isolating a percentage of their user base is acceptable. That is a decision for the developer to make.

For a complete archive of browsers to test with, visit evolt.org. They have almost every browser in existence available for download.