🚀 go-pugleaf

 In article <833197612_OASIS_@irana.demon.co.uk>
  gen@irana.demon.co.uk (gen@irana.demon.co.uk) said:-

> In article <23948228_Beta04_OASIS_NEWS@nest.demon.co.uk>
>  dbrewood@nest.demon.co.uk (Daron Brewood) said:-
>
>>>1) BAD Packet errors:- these occur when downloading news, I have tried all three
>>>settings of packet size, and also SLIP instead of PPP but errors still happen.
>>
>>This is the one big bug that is killing Dave, he has coded to the RFC's 
>>pricicely, but can;t figure out why this one is generated.
>
>NOS seems to be able to recover better, it does stop from time to time but it
>adjusts. I've seen such things as changing MSS size! Rtt is assesed constantly
>too, I presume... BTW, what is backoff timer?

The backoff timer is used for errors. There's no point in a socket 
being madly kicked by one end if the other end is not responding. The 
backoff timer in Nos rises exponentially. When the timer next reaches 
zero the socket is kicked - if a response is received then "the 
backoff" is cleared to zero otherwise the delay is increased (Nos uses 
new_val = old_val << 1). Obviously this formula can quickly raise the 
delay into minutes.

If you are certain the cause of the delay has gone (typically modem 
line noise) then a manual kick can bring a socket back into life very 
much earlier than it would do on its own. In most other cases you will 
just make the problem worse. Use "tcp kick <TCP_address>" for a manual 
kick.

"tcp reset <TCP_address>" can usefully be employed to abort a 
connection. My most common need for this is when a large email has 
accidentally gone direct to a slow destination. Nos will then try to 
connect via the gateway where I intended in the first place.


__

Guy Harrison

guy@swampdog.demon.co.uk