You couldn't compile it from source since there isn't any. Picasa uses some free software (wine for instance), but isn't free software itself. However, there's an ebuild already:
http://bugs.gentoo.org/show_bug.cgi?id=134376
Source code unavailability was claimed to be good for security. Crackers can't read our code, so they won't find holes. It's harder to find holes by trial and error than by reading and analyzing the code. On the other hand, a lot of Windows source code has leaked into p2p networks some time ago. So crackers have a lot of reading. How does Microsoft security team feel about this?
Why not calculate a simple regression? I know that it doesn't follow the way you normally calculate statistics. But let's say one task for one system is one observation. Then you would get something like this:
Call: lm(formula = secs ~ fs, data = F)
Residuals:
Min 1Q Median 3Q Max -21.554 -14.636 -7.982 14.958 97.796
Residual standard error: 19.71 on 120 degrees of freedom Multiple R-Squared: 0.01424, Adjusted R-squared: -0.02683 F-statistic: 0.3467 on 5 and 120 DF, p-value: 0.8835
So the "estimate" column shows the relation between the filesystem and overall mean. Ext3 with -0.02 is just average. JFS with -3.39 is the fastest. XFS with -0.07 is just a little bit faster than EXT3.
This "analysis" assumes that each task is equally important, which is not true. So I think the best test would be to log all the disk activity and log it, so it would look like:
The normal disk activity would get logged. Then one would just spend one day with each filesystem. Then one would make a regression that would show a real (or at least close to real) estimation of each filesystem performance.
Slashdot Mirror
You couldn't compile it from source since there isn't any. Picasa uses some free software (wine for instance), but isn't free software itself. However, there's an ebuild already: http://bugs.gentoo.org/show_bug.cgi?id=134376
What about the direction of the coil -- they positioned it parallel to the power line. Maybe it should be perpendicular?
Source code unavailability was claimed to be good for security. Crackers can't read our code, so they won't find holes. It's harder to find holes by trial and error than by reading and analyzing the code. On the other hand, a lot of Windows source code has leaked into p2p networks some time ago. So crackers have a lot of reading. How does Microsoft security team feel about this?
This "analysis" assumes that each task is equally important, which is not true. So I think the best test would be to log all the disk activity and log it, so it would look like:The normal disk activity would get logged. Then one would just spend one day with each filesystem. Then one would make a regression that would show a real (or at least close to real) estimation of each filesystem performance.