🚀 go-pugleaf

In article <5ctjua$jq5@usenet78.supernews.com>,
	webmaster@arselectronica.com (Michael Shavit) writes:

> sich ja auch nicht mit Zunahme der Lichtstaerke (Lux). Nur hier kommt der Clou 

ping_delay == time to transmit the packet
              + time to transmit the answer
              + computing/relaying time

So if you double your line's speed, you should see a quite good
increase for big packets. For small packets, or if the bundling
allocation algorithm is based on a treshold (ie it does not send one
byte here, one byte here, but instead blocks, which are probably
easier to handle), small packets will be sent slowly, ie probably one
one line. If someone knows the ``partitionning'' algorithm for
bundling, it would be nice to tell it.

You should really use UNIX ``bing'' (available on ftp.ibp.fr)
which uses different packet sizes and on that basis evaluate
the bandwidth between two points.

Or, if you don't have luck (ie !UNIX), try to do some pings with
bigger packet sizes (ping -c SIZE) and/or some FTPs.

Asymptotically, for big packets, the rate should be doubled
with ISDN bundling.

PS: Don't forget that WWW sessions (which seem to be quite popular
    nowadays) use TCP sessions (opened and closed usually for each remote
    object they download; HTTP 2.0 excluded), which means THREE (small)
    packets just to open the session (-> <- ->).