Right. In the PETS paper they were looking at unique identifiers over a
month. Your way of looking at hourly numbers is realistic because of
churn. CENO doesn't mention how they measured anywhere.
On 2015-10-25 01:05, Steve Dougherty wrote:
We're not using Google analytics; that bug is outdated since the new
site
launched a few days ago. We are now using piwik webserver log analysis.
I don't know where those other estimates came from. It'll be
unsurprising
that my numbers sound the most reasonable to me. The stats site goes
into
detail on what exactly the estimates are of and how they're generated;
you
can compare that to the PETS and CENO techniques. I'll be interested to
learn how they differ - could you report back if you find anything? I'd
actually expect at least one of those to be the total number of unique
identifiers observed over some period, which does not reflect network
size
as far as number of nodes regularly online.
I'm not sure how accuracy might be improved. As mentioned in the
explanation section of my stats page, weighting by the uptime
percentage
included with each probe identifier might help, but it's not especially
clear what sort of weighting is appropriate.
- Steve
On Sat, Oct 24, 2015, 11:39 PM <salutarydiacritica...@ruggedinbox.com>
wrote:
For Pete's sake don't run Google analytics on your site. I saw a
ticket
for it on bug tracker. Piwik is an open source analytics system you
can
try instead.
Speaking of analytics, I am trying to learn the network size. Steve's
site reports 10K nodes, the 2014 PETS paper puts it around 60K nodes
after harvesting node information and CENO says theres 100K nodes.
Which
do I believe and how can there be more accuracy?
_______________________________________________
Devl mailing list
Devl@freenetproject.org
https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
_______________________________________________
Devl mailing list
Devl@freenetproject.org
https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
_______________________________________________
Devl mailing list
Devl@freenetproject.org
https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
