Date: 24 Sep 2001 13:00:58 -0700 From: swear@blarg.net (Gary W. Swearingen) To: Joe Abley <jabley@automagic.org> Cc: Juha Saarinen <juha@saarinen.org>, "'Andrew Reilly'" <areilly@bigpond.net.au>, freebsd-stable@FreeBSD.ORG Subject: Re: 127/8 continued Message-ID: <swr8swwe85.8sw@localhost.localdomain> In-Reply-To: <20010924070102.I4205@buffoon.automagic.org> References: <20010924160936.A10863@gurney.reilly.home> <00e001c144c8$c33bf900$0a01a8c0@den2> <20010924070102.I4205@buffoon.automagic.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Joe Abley <jabley@automagic.org> writes: > RFC 1122, "Requirements for Internet Hosts -- Communication Layers" > provides guidance for the interpretation of any address within > 127/8 -- it says such addresses are for use as "internal host loopback > addresses". RFC 1122 is STD 3, an Official Internet Protocol Standard, > and hence is worth complying with. Are IANA/IETF/Internet standards EVER applicable to what goes on inside our computers? Or just to the data crossing our Internet interfaces? (Not rhetorical - I'm wondering.) > RFC 1122 does not state that "every possible address within 127/8 > must be treated as though it is a configured loopback address", > and to interpret it as such is bizarre and counter-intuitive. Be nice; that's a tough thing to say convincingly, intuitions being what they are. I don't think the RFC should even be addressing the issue, but if FreeBSD goes along and uses 127/8 address as "internal host loopback addresses" (as quoted above), it seems fair (and intuitive) to me to say that those addresses should have no other use and it would both keep them inside the host and save people the effort of configuring separate loopback aliases if they need them. FreeBSD does 127.0.0.1 for free; it might as well do them all, especially since it's no extra effort (more than blackholing them). > Installing a null covering route for 127/8 with the blackhole bit > set seems a good way of preventing addresses with a destination > within 127/8 from being sent out on a non-loopback interface, without > resorting to nasty hacks which make address handling on the loopback > interface different to every other interface. It is also consistent > with the robustness principle. I don't see need for any hack. (None beyond the need to mess with this stuff at all. I think this stuff would be better hidden and out of the routing table listings, boot scripts, etc.) Having loopback addresses automatically loop back shouldn't suprise anyone. Now, maybe I'd need to modify my opinion if the lo(4) man page indicated that the "pseudo-device loop" KERNCONF line had "[count]" in which case it probably wouldn't make sense to have all loopback addresses looping back to the same lo0 by default. (I saw some code which looked like it supported multiple "lo#"s, but the man page and LINT give no hint of it.) > route add 127.0.0.0 -netmask 255.0.0.0 -iface lo0 -blackhole I sure don't see how anyone could get that from the "route" man page. I'd write a PR+patch on it if I had a clue. I think it needs to say that the gateway may be specified as "interface" (eg, "lo0") when the "-iface" modifier is used (the page implies that's only valid with "-interface"). Is the placement of modifiers critical? The man page's syntax lines looks wrong, as does the discussion of the specification of netmasks. > But, whatever. This is hardly a monumental requirement worth bickering > over. But apparently worthwhile to Joe, me, and others, if not FreeBSD. Looks to me like a (sometime-counter-productive ;-) attempt to discourage other opinions. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?swr8swwe85.8sw>