On 18/12/06, David Gerard <dgerard(a)gmail.com> wrote: I remember them as well; there were some on Kohl, I think, in home directories. It's possible those got lost/disabled somehow when we had that whole NFS mount business? Rob Church

On 12/18/06, David Gerard <dgerard(a)gmail.com> wrote: [snip] You're giving too much credibility to the alexa data. There are sites which are ranked somewhat lower than us which I know for a fact are processing more req/s. It is indisputable, however, that we are one of the more visited sites around.

On 18/12/06, Rob Church <robchur(a)gmail.com> wrote: They'll be reading this thread and want the real numbers? ;-) - d.

On 12/18/06, David Gerard <dgerard(a)gmail.com> wrote: Three cheers for the self-selecting sample. On 12/18/06, David Gerard <dgerard(a)gmail.com> wrote: Well, the link has been given... and the same data has been previously provided to other donors, it's just that it never made it onto the pages. The graphs themselves are a bit hard to read because the daily volatility swamps the short term growth... If only I there were stocks which I could trade that followed our traffic patterns. Keep in mind that roughly 2/3 of the overall requests are images, so a correction must be applied to convert req/s to approximate page views/src.

On 18/12/06, Gregory Maxwell <gmaxwell(a)gmail.com> wrote: That's a pity - 70,000 users is really quite a high-resolution sample [*]; it's a pity their sample selection is evidently crappy. The traffic stats yearly graph makes that clear. 1 gigabit/second sustained, doubling in the last six months ... if we could just slow it down to Moore's Law rates ... [*] http://en.wikipedia.org/wiki/Margin_of_error - in a result which many statistics students find surprising, the margin of error depends *only* on the number of samples taken, not on the size of the sampled population. Most political polls, for example, sample about 2000 people; sampling 70k is a *lot* of people. - d.