M P H
Creative Commons License photo credit: kyz

Last night, after working for several hours to boost the load-time of pages on the popular economics blog MarginalRevolution, I was at a loss for what else I could do to boost the site’s performance as indicated by Google’s PageSpeed and Yahoo’s YSlow evaluation tools.  I had only managed to boost its bad score from 71 to 78—still in the C range, as these work like letter grades—by enabling aggressive page caching and consolidating the site’s CSS and JavaScript files.

Then I removed the Facebook and Twitter buttons provided by AddThis, the social media sharing service, and the score shot up from 78 to 93.  The site falls short of a perfect score because of the remaining calls to services like SiteMeter and BlogAds, but an A is still an A, and new load time of around one second wasn’t too shabby.

From what I can tell, AddThis places the blame for this on Facebook and Twitter.  It seems both social networking sites have APIs that leave much to be desired, or at least servers that can’t keep up with all the requests for post counts.

But is it really fair for AddThis to blame Facebook and Twitter?  Couldn’t calls be made once every few minutes to the API to get a new count, rather than every time a page is loaded?  Couldn’t caching happen somewhere in that chain?  Real-time social networking information hardly seems worth slowing a site down by several seconds per page load, while delayed stats and fast pages seems like an entirely reasonable compromise—it would certainly be better for AddThis, Facebook, and Twitter than no buttons at all.

So, you’ll notice the comment-count buttons are now absent from MR, but thanks to the incredible popularity of the site, I don’t think readers will be wondering if anyone’s still passing links around.