[pskmail] Protocol enhancement in pskmail...

  • From: Rein Couperus <rein@xxxxxxxxxxxx>
  • To: pskmail@xxxxxxxxxxxxx
  • Date: Sun, 08 Mar 2009 20:37:35 +0100

After 5 years of testing pskmail, I had the chance to simplify the protocol and 
make data transfer more efficient.
The original protocol, after an idea of K9PS, was based on a fully symmetrical 
geometry.
Both server and client have the same rights, and both sides contain timers 
which take 
action in their own pace, resulting in quite some overlapping transmissions, 
which 
occurred when one of the stations were not hearing each other.

The firsts step (2 years ago) was to introduce a software DCD, which prevented 
transmission 
when another station was sending.
Step 2 was to synchronize poll sequences... the client polls in seconds 0-4, 
the server polls in seconds 5-9.
The problem here is that a lot of clocks go wrong... it is ok when you have a 
gps, but 
this has given problems to some users....
Still this meant an enormous increase in throughput on mediocre links.

I have now taken step 3. which means the client does not know about timing 
anymore, the server controls 
all the traffic. Now (with the new java client) there are no more overlaps in 
transmission. The client 
waits for the status block of the server. If that is damaged, the client just 
waits for the poll.
This looks too simple to be true, but it works,,,,

The new protocol is compatible with the present servers. They don't know the 
client 
has lost its timers, and are glad nobody is talking in between.....

The link has become a lot more robust as well... one session this afternoon 
lived through 
an RTTY qso which suddenly came up on frequency (9+20dB)...

73,

Rein PA0R
.


-- 
http://pa0r.blogspirit.com

Other related posts:

  • » [pskmail] Protocol enhancement in pskmail... - Rein Couperus