all zeroes/all ones used in host IP's...

[Magnus Danielson]

From: "Michael H. Warfield" <mhw en wittsend.com>
Subject: Re: all zeroes/all ones used in host IP's...
Date: Fri, 28 Jan 2000 23:16:32 -0500

Hi!

> On Sat, Jan 29, 2000 at 02:53:57AM +0100, Magnus Danielson wrote:
> 
> > Please enligthen me if you can find an error in my reasoning, all under the
> > assumption of an non-CIDR case.
> 
> 	Your reasoning is not at fault, it's your premise.  It's just that
> a lot of us (I won't say all of us) who actually have responsibilities for
> larger address spaces (I have a /16 plus ISS has a /19, two /20s and a /21
> in the corporation), and have to run them, consider the non-CIDR case to be
> either a "legacy" situation, an annoyance at best, or an excuse by vendors
> for why their routers #$@$#@ up at worse.  It really ends up boiling down
> to netmask and netmask and netmask and nothing much else.  Class A, Class B,
> and Class C degenerate into nothing more than the DEFAULT initial netmasks
> (/8, /16, and /14 respectively) to be modified by appropriate routing
> tables and configurations.  I have lots of networks where I see the
> "illegal" one bit wide subnet masks.  Those restrictions from RFC 1122 are
> largely meaningless in a CIDR world.

No, you have just not understood what I said. I know all the crappiness of
Class A, B and C. I know CIDR is much better and I have been encouraging it's
propper use. I know the this is all legacy and all that. However, what I have
been pointing out is that if your network is NOT a CIDR activated network these
rules are still standard from the IETF/IESG point of view. Yes, they are
broken by todays defacto standards, but if you take the time and investigate
things, you will find that these restrictions are still in power if you are
going to follow the standards to the letter. I have also said that this is not
a big issue for the linux kernel, since in practice almost all networks (to the
best of my knowledge) is being run as CIDR networks and for these purposes will
these rules have no applicability.

For the latest status on standard level of various RFCs, please go to the
RFC Editor (http://www.rfc-editor.org/) and read for yourself.

> 	To paraphrase uncounted paraphrases of a paraphrase of an
> expression...
> 
> 	There are two classes of networks...
> 
> 	Those that are CIDR and those that will be CIDR.  The "non-CIDR
> case" is a non-op in the real world of the current internet.
>
> 	Hmmm...  That may be a BIT strong...  Non-CIDR can interact with
> the core internet.  It's not a non-op per se...  It's much more of a pain
> in the fanny than a non-op.

Yes, non-CIDR is just a local pain but as you get outside people will not
really notice.

What people seems to not properly understand is the separation between defacto
standard (in this case all is CIDR) and actual standard (non-CIDR and CIDR on
the way to become standard). So, if someone to be "right" according to what is
being standard then you must be able to separate these issues. Regardless of
how much we hate it, non-CIDR is still standard and if some network admin still
feels he just run non-CIDR and subneting, in this case these rules would apply.
However, nobody will kill you or even slap your face for not complying.

So, to answer the original question, sure Linux does not treat these cases and
has no intention of doing so since it doesn't well match todays and future
reality. This is a sidestep which seems widely accepted today.

Cheers,
Magnus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo en vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/