God, yes. It frustrated me so much that I ended up just copying the data into JS arrays and generating my own charts using the Google Visualization API.

http://dl.dropbox.com/u/25782/sourceforge.html

Pretty eye-opening. What's clear is that the 'Other' category brings so much uncertainty to the data that it renders the data virtually useless.

I found stacked area to work really well:

http://dl.dropbox.com/u/6353242/hn/sourceforge-area-normaliz...

You can get a real sense for how large the 'other' datapoint is by looking at non-normalized view as well:

http://dl.dropbox.com/u/6353242/hn/sourceforge-area.html

Apparently much of the "other" data are non-browser tools. E.g., curl and wget.

https://twitter.com/#!/rbowen/status/137305778875936771

Follow up conversation at: http://www.bradlanders.com/2011/11/17/visualizing-sourceforg...

That's an awesome chart to visualize

Indeed. A single chart with time on the horizontal axis, downloads on the vertical axis, and one line per operating system would have been far more comprehensible.

Instead we get time inverted, bar graphs, and no way to visually compare the operating systems.

It also isn't very good to have an "other" category of ~60M averaged when your three "main" are ~60M, ~4M and ~3M.

Terrible article.

Apparently, the unknown represent the auto updates right through the application itself, but sourced from SF

https://twitter.com/#!/rbowen/status/137305778875936771

Rather.

It took me a while to figure out the dates were MM/DD/YY. Even then, why is the day included? Are these per-day, and if so, why are they making sweeping generalisations about all of April? If they're just point markers why aren't they in line, and why are the bars separated?

It's great they managed to provide the full stats - in a completely uncopyable format. Let me break out tessaract. http://i.imgur.com/VDtUp.png http://i.imgur.com/nuKxl.png for (only marginally) more helpful visualisations

EDIT: warpdude's ones four hours ago and two comments down look nicer