Steve Souders has studied @font-face, and comes to the conclusion it's a performance hog because IE waits for the font to be downloaded before showing the page, Safari hides the text until the font is downloaded, while Firefox initially shows the page in the default font and then abruptly changes the font when the download is done.
None of these solutions are optimal. The real problem is that they're the only solutions available, so even browser progress won't solve the fundamental issues.
Performance
| Permalink
Another browser speed test, this time for the client/server browsers Opera Mini, Bolt, Skyfire and UCWEB. Opera Mini (barely) wins.
Opera Mobile/Mini, Performance
| Permalink
A browser performance test on mobile. Surprisingly, the S60 WebKit (five versions tested) is faster than either iPhone or Android.
The only thing I'm missing here is a reference to S60's extremely aggressive caching. Still, I might use the methodology described here myself later on.
Android, Performance, S60 WebKit, iPhone
| Permalink
Stefan Kolb continues his series of tests of JavaScript libraries on mobile phones. This time he did the TaskSpeed tests on ten Nokia S60 phones.
Conclusion: Dojo again the fastest library; this time Prototype is the slowest.
Libraries, Performance, W3C Widgets
| Permalink
Steve Souders discusses situations in which it might be useful to spread your assets over multiple (sub)domains, even though that will lead to extra DNS lookups.
Performance
| Permalink
Stefan Kolb, one of my co-workers at Vodafone, has conducted a selector performance test for seven JavaScript library versions in the Vodafone Widget Manager, which runs Opera Mobile, on ten different Symbian S60 phones.
For now it is clearly visible that some frameworks perform better than others in terms of DOM selection. By far the slowest framework in my tests across all devices was the YUI v2.7.0 framework. The fastest frameworks were the two version of the Dojo framework, with version 1.3.0 performing slightly better than version 1.2.3.
It is also clear from the results that the performance depends on the mobile device. The Nokia N73 was the slowest phone, no matter which framework was tested on that device. The fastest phone was the Nokia E66, closely followed by the Nokia N85.
Hopefully, the tested - well established - web frameworks will soon be optimized to perform better on mobile phones. After that, I am sure, they will be of great value for the mobile widget developer, just like they are for web developers today.
Obviously, we need many, many more performance tests before we can say which library is "best" on mobile phones. Still, today we've made a start.
Libraries, Performance, W3C Widgets
| Permalink
Steve explains how script tags block other downloads in older browsers, although newer ones (IE8, Saf4 and Chrome2) do not block downloads, though they still may block rendering. Finally he gives a few possibilities for loading scripts in a non-blocking way.
Performance
| Permalink
Steve Souders explains why you should not use @import.
Performance
| Permalink
Nicole Sullivan about repaints and reflows and what the performance cost is. She adds a few interesting rules to optimise your CSS change performance.
Performance
| Permalink
Yahoo! had already published 14 guidelines for keeping your site performance down. Today they announced 20 more guidelines. I haven't yet studied them in detail, but I'm sure they're just as important as the first batch.
Performance
| Permalink
John talks about performance testing JavaScript libraries and how not to go about it.
JavaScript, Performance, Tests, Theory
| Permalink