On Thu, 2006-05-04 at 07:38 +0200, Daniel Stenberg wrote: > Yeps. In my quick and naive zip/7zip compression tests I did a number of > months ago, the typical gain for a rockbox.zip was in fact more than 30%.
Since rockbox.zip tends to be pretty small I imagine it's skewing the results a bit (30% of 2MB is only 600k). > Since we're talking about packages that are downloaded thousands of times > *every day* we would also cut down on bandwidth for the server(s) hosting the > downloads. (If we also count the daily builds.) It seems to me the smarter solution would be to separate the binary & lang components from the WPS/themes/fonts parts (as was discussed), rather than just switching to 7z. That said, I'm not really *against* 7z since it's LGPL and available for all major OSes. bk
