Re: Client ID?

Karl J. Runge (runge@crl.com)
Thu, 8 May 1997 14:58:11 -0700


On Thu, 08 May 97, "Dean Mills" <dmills@cablelan.net> wrote:
> well, still nice to see OS/2 doing well :) Wish those stats were
> available for DESChall as well, I know a few advocates who wouldn't
> stand for anything but #1 position for their respective OS's. Any
> chance on these stats being available?

Hi, I'm the one working on that, I've signed a NDA with Rocke to protect
the anonymous hosts and now have a prototype that parses Rocke's RCV logs.
Works OK on a small (~ 3-5%) subset of all the logs he sent me.
I'm about 1/2 way done with the HTML spitter. After I access all of the
logs and pass things by Rocke I'll have a prototype to show you all.

Some notes:

The main focus of my programs is just to get *any* improvements to
the statistics. They are not really accurate. Too many problems
with UDP resends to get stats 100% correct up front. Good news is
for the 2 full days of logs Rocke sent, my log "model" agreed with
Rocke's numbers to within 1%. Close enough for government work!

It runs slowly and sucks up alot of RAM here on my tiny 486's (Only
16MB RAM avail). Furthermore, I have a modem connection and not much
disk space and the log files are LARGE. I have worked out an
arrangement to use another machine, but that may fall thru.
Anyone out there with a P5-P6, >32MB RAM >700MB HD free, fast net
connection I can run this on daily? You may have to sign an NDA
with Rocke, but I am not sure.

To, finally, answer your question, my parser saves what it can of the
OS Platform. I haven't written the tallyier and HTML spitter for that
yet. I want to have Platform information but it may not be super
accurate, the problem is that info has to be traced back in time to
when the client was started (which may have been many days ago, IP#'s
PORT#'s changing, etc.). I may make some optimistic assumptions here ;-)

I'm aiming for releasing preliminary stats this weekend. Wish me luck!

Karl