Histograms Of Furthest States Reached By Hello Link-Clickers In Trysts

gistfile1.txt

Of rooms in which there were ever 2 people simultaneously, here are the
furthest states reached by the link-clicker.

We'll have state info for the built-in client starting in v40. "Further"
states are toward the bottom of the histograms. Histograms are out of scale
with each other.


2015-05-01:

    leave  10
  refresh  4
     join ********** 223
  waiting  0
 starting  5
receiving  14
  sending *** 78
 sendrecv ************************************************************************************ 1838 


2015-05-11:

    leave  8
  refresh * 40
     join ** 52
  waiting  1
 starting  2
receiving  17
  sending ** 53
 sendrecv ******************************************************************************************* 1934 


2015-05-21:

    leave  9
  refresh  2
     join ** 51
  waiting  3
 starting  4
receiving * 22
  sending ** 51
 sendrecv ******************************************************************************************** 1867 


2015-05-31:

    leave  2
  refresh  0
     join ** 46
  waiting  0
 starting  0
receiving * 17
  sending ** 41
 sendrecv ********************************************************************************************* 1555 


# Observations:
#
# * Most rooms never see 2 people meet: 20K lonely rooms vs. 1500 meeting ones.
# * There are many sessions consisting of nothing by Refresh events,
#   generally a mix of clickers and built-ins. Where are the joins? On a
#   previous day? It would be interesting to see if these happen near the
#   beginnings of days.
# * There are some sessions in which leaves happen without symmetric joins.
#   See if these occur near the beginning of days. Otherwise, I would expect
#   at least Refreshes every 5 minutes.
# * These numbers may be a little high because we're assuming all
#   link-clickers are the same link-clicker. When we start logging sessionID,
#   we can start distinguishing them. (hostname is the IP of the server, not
#   of the client.)
# * These numbers may be a little low because we don't yet notice timeouts
#   (client crashes, etc.), making the denominator falsely high.