[ntp:questions] Re: Multicast TTL and FreeBSD trouble

[David L. Mills]

Roger,

I had a hand in the original spec rfc791, which interprets the TTL field 
values in seconds, with the understanding that every router decrements 
by one and in addition decrements once each second, assuming the grams 
survive that long. The only implementation that actually did both of 
those things to my knowledge was the fuzzball. In the bad old 
ARPAnet/NSFnet days, grams could survive that long on a router queue.

Dave

wa6zvp at gmail.com wrote:
> I'm not saying that there is an error in the code, per se, but
> somewhere when the packet is assembled for putting on the wire, the TTL
> is mis-handled.  On a RH linux machine, the byte is correct on the
> wire, implying that it must be in the interface to the IP stack or some
> such place.
> 
> For NTP use, most folks would want the multicast packets to stay within
> the boundaries of their own local network, which means a TTL in the 1-3
> range is probably appropriate.
> 
> As Danny pointed out, TTL for v4 used to really be time (was it
> seconds?) but has been used as a router hop count for a long time.
> This is how its defined for v6.
> 
>          Roger
>