Re: Should we share?
- To: email@example.com
- Subject: Re: Should we share?
- From: Phil Karn <firstname.lastname@example.org>
- Date: Wed, 7 Dec 1994 18:44:50 -0800
- Cc: email@example.com
- In-reply-to: <199411281629.IAA21769@nothing.ucsd.edu> (firstname.lastname@example.org)
Am just now catching up with this thread.
>But this begs the question:
>IP routing should NOT be used instead of routing at the AX.25 level.
>This is one of the grossest lacks in the KA9Q package (and everything
>derived from it, as far as I know).
>I divided up the network 44 address space for ADMINISTRATIVE convenience,
>NOT for routing purposes. Few if any of the existing amateur networks
>physically follow convenient subnetting boundaries. Localized subnets
>should rarely be necessary.
As the author of the aforementioned gross hack, I happen to agree 100%
with Brian. While I do think my scheme for variable length subnet IP
routing entries was a good idea that IMHO took far too long to be
reinvented in the real Internet as CIDR, it is by no means a proper
substitute for some sort of internal routing within the subnetwork
level of AX.25.
What we need, basically, is "NET/ROM done right". Build a ham-specific
network protocol that uses callsigns as addresses, but use a decent
platform (rather than a TNC-2) and a decent routing algorithm (like
SPF instead of DV). Then you'll be able to build metropolitan area networks
that are fully connected as far as IP (or anything else) is concerned,
and things would be a lot easier.