Gnupg good for big groups?

[Robert J. Hansen]

On 8/9/2010 1:48 PM, MFPA wrote:
> I worded that very badly. Rather than throwing away data, I was trying
> to suggest looking at it in more detail. As a whole from day 1 to day
> n, the proportion not encrypted to the key was x. Fine as far as it
> goes, but did the figures confirm or refute my expectation that the
> proportion would be higher in the first few days and then decline to a
> steady level?

You would have to ask Paul.  I suspect, though, that with only a
low-thirtysomething number of nodes and a total number of messages in
the neighborhood of six hundred, that there's not much confidence to be
had in any trend.

Gross behaviors (the combinatoric explosion of edges as new nodes enter
the graph, churn in the fringes, etc.) are fairly easy to recognize in
even small data sets.  Subtle behaviors (figuring out precisely what the
problem die-off is) are difficult to discover and require some pretty
sophisticated knowledge of statistics -- far beyond my own capabilities.