[e2e] tcp keepalive 2 hour default

[Craig Partridge]

In message <001f01c0c8e4$f351b1d0$7203010a at ennovatenetworks.com>, "tom worster"
 writes:

>does anyone here know the historical rationale for the two
>hour minimum on the default tcp keep-alive specified in 
>rfc1122?

I did a quick look through the host requirements archive and the basic
answer is that it was a compromise...

The key issue is that short keep-alives are actually anti-robustness
(that is, a path change, or brief network outage may cause a TCP connection
that is serving its data just fine to fail - think TELNET or X-Windows
and long pauses between user key strokes).

Long keep alives don't very effectively help people who want to real
TCP connections periodically.

After some (not clear how much) discussion, we landed at the two hour number.

Craig