I got several requests emailed to me personally from other network
operators who also wanted this information, so I'm posting a summary
back to the list. If you feel this information is beneath you, by all
means stop reading right now!
Basically, the reason why I had to go out of my way to enable the
subnet where the network portion of the last octet was all-zeroes or
all-ones is a relic from Ancient Times (read: about 18 months ago),
when classful networking was everywhere, and RIPv1 walked its hallowed
halls. RIPv1 does not send subnet information in its route
announcements, so the subnet that contained the Class C network and
broadcast address were off-limits.
While many routers come this way by default (a situation which one
respondant described as "abominable"), everyone seemed to be in
agreement that any router that did not totally suck would let me turn
this off. A few others added to this that they felt my Bay Networks
router *did* totally suck. As proof that this isn't the case (at
least not this time), the problem ended up being completely unrelated
to the subnetting. But I was still curious.
On a Cisco, the command is "ip subnet-zero" and "ip classless". On a
Bay, the "Enable zero subnet" button in Site Manager does the trick.
As backup ammo against future tech support incidents, several people
pointed out RFC 1878 and 1519, which lays all this out.
With regards to reverse DNS, several people pointed me to the advice
in RFC 2137. It's just a few pages, and sums up the situation better
than I could.
Thanks to the numerous people who responded by private email,