tcp/ip assumptions

[Mikus Grinbergs]

I have a G1G1, which communicates via a local LAN to a "relay" 
system, which communicates to the internet.  The server facilities 
(*just* for the XO - not needed by the "regular" systems on my LAN) 
I've now set up in the "relay" station are minimal (e.g., for DNS). 
  The result is that many XO requests are not fulfilled by my 
"relay" system (for instance, a separate dialog may be needed - 
between the "relay" system and a *real* server out on the internet).

I was looking at a trace of the packets on my local LAN.  In the 
case of DNS, the XO issues three "Type 28" requests (which my 
minimal "relay" station does not support), before issuing a "Type 
01", to which it eventually does get an answer.  In the case of NTP, 
the XO issues "scattershot" requests to all server addresses it was 
able to extract [but receives no responses, because it tries to 
contact them directly, rather than going through the 'proxy' 
function in my "relay" system].


My conclusion:  The tcp/ip function in the XO makes a number of 
assumptions as to the type (and timeliness) of the external SERVICES 
it expects to have been provided.  I would have been happier if I 
had known about these beforehand, rather than having to discover 
what does or does_not work in the environment I currently have. 
[Might some setups in a target country be as minimal as mine?]

mikus