Brett Mueller wrote:
>
> Hi Tom,
>
> On Fri, 19 Feb 1999, Thomas Sailer wrote:
> > First, it's not a good idea to use datagram mode to transport IP.
>
> I am curious about this statement. Is it because of not being able to
This has been discussed to death already, but here's the short list:
- Flow control (as you guessed)
- Cumulated packet error rate over multiple links is far too high
- Can't do segmentation (i.e. long IP packets) reasonably over DG.
> real-world), it wouldn't make sense to send TCP (with guaranteed delivery)
> over IP (which does not guarantee delivery) over a connected-AX.25 circuit
> (which would guarantee delivery), without some very strong reasons. From
Yes it does. AX.25 CRC gives you hop to hop data integrity,
and TCP gives you end to end data integrity (i.e. protects from
malfunctioning switches in between, which is can be a real source of
data integrity problems).
TCP btw. doesn't work well with packet loss rates more than
a few %, because it treats every packet loss as congestion and
backs off.
IP btw. gives you independence from AX.25 as transport technology
(i.e. you can have segments of data transmission technology
different from radio, i.e. wires in between), but if that
is an advantage or disadvantage is actually quite a religious
issue.
TCP/IP overhead is almost negligible if VJ compression is used.
Tom
