Packages and sources for R are distributed and hosted at a network of volunteer-run FTP sites, collectively known as CRAN mirrors. There are currently 69 CRAN mirrors worldwide, and the CRAN statistics page helps you find one that's working. But you'll get better performance when downloading packages -- and help share the load -- if you choose a geographically local mirror.
Barry Rowlingson has created a couple of tools to help you do just that. The first is Cranography, an interactive map showing all of the CRAN mirrors around the world. Barry used the geocode geonames package to determine the locations of the mirrors (with some manual adjustments), and the map itself was built using OpenLayers. Click on the screenshot below to try it out.
So that's how you can find the closest CRAN mirror, but is it the fastest? Barry offers a solution there, too: the cranometer function. It estimates the speed of each CRAN mirror by measuring the time it takes to download the NEWS file. It's a rough-and-ready estimator (be sure to check the caveats) but it's certainly an interesting application of the internet-capable functions in R.
Barry Rowlingson: Cranography
It seems that the Cranography web site is off-line right now! :'-(
Posted by: Paolo | August 05, 2009 at 01:57
Barry let me know I picked the one day the servers were down for maintenance to link to his page. It's back up now.
Posted by: David Smith | August 05, 2009 at 12:02