The Issue with Client-Side
As “web apps” become more and more prevalent, it seems that there is once again a demand for client-side compute power. Web browsing used to be a fairly processor-independent activity, but inefficient Javascript engines, excessive Flash, and badly designed dynamic sites have change that. Web sites can tax your browser and, unless you have a high-powered machine, can interfere with actually using the application.
Early builds of Twitter’s new site were one example, with Firefox choking on its excesses. The old Twitter was a fairly simple site, but the new one added so many dynamic layers and alpha blending effects that it bogged down the browser. Improvements to Firefox and Twitter have since resolved this problem, but it’s clear that efficient Javascript in these web applications can make or break a site.
Another example is the new console for Amazon Web Services. It’s pretty and works great, but would be so much more responsive if it were an actual desktop application. In the browser, it’s sluggish and has even locked up Firefox a few times. It reminds me of the bad old days of Java.
These are just a few examples of the direction the Web seems to be heading. It’s no longer appropriate to think of web applications as light-weight versions of a desktop app, as they can eat up more CPU cycles with just a few ill-formed loops. CPU makers certainly do not need to fear being left behind with these trends looming.