On Aug 16, 2010, at 8:37 PM, Hemant Singh (shemant) wrote: > -----Original Message----- > From: Jared Mauch [mailto:ja...@puck.nether.net] > Sent: Monday, August 16, 2010 12:50 PM > To: Hemant Singh (shemant) > Cc: Randy Bush; ipv6 deployment prevention; Wes Beebee (wbeebee) > Subject: Re: Router redirects in Node Requirements document > >> So it's a non-issue and they don't need the redirects in this case it > seems. > > No, it is an issue. Did you see my email for the detailed use case that > two neighboring homes with two different bridged cable modems have to > talk to each other throught the SP first hop router (CMTS). But the > redirect example I gave was for two hosts behind the same cable modem in > the same home - when one host sent traffic to the other via the SP first > hop router, the router HAD to send a Redirect!
So, you are in favor of wasting the RF space by making the packet have increased latency through a remote device not in-home for in-home communication? I knew that docsis/cmts were a bit strange, but you really expect the hosts to be doing this? Is there a cable provider doing such crazy routing activity today? I doubt it. It sounds like you support: Having two hosts on the same LAN (in-home) sending all their traffic via the CMTS increasing the limited upstream/downstream network utilization vs having the hosts have properly scoped subnet masks. I support: requiring a host that has an IPv6 address to explicitly have some sort of mask (eg /48 /64 /96 ALL examples) configured. Having the host think it's an island unto itself (/128) and relying upon the grace of technology to direct it places additional load on the actual network devices. > >> Just because someone can poorly manage their network doesn't mean you > should be advocating it. I'm not sure I see any operational "win" here. > If you step outside the CMTS world, you typically have >better defined > subnets/masks on devices running IOS/XR/XE/JunOS/JunOS-E/NxOS even when > the device is acting as a DHCP server doing ND RA (vs lots of diverse > subnets coming in; biz vs consumer vs voip). > > Even outside of the CMTS world, soon as one has a LAN network with more > than one router, one can easily see a need for Redirect - google for so > many examples of LAN networks where Redirect is used. Also, the > examples for other OS and vendors and DHCP server in the same router is > all moot - the same router sends the ND RA and the DHCPv6 server will > not send the prefix len in DHCPv6 responses. I did not realize you spoke for google. I will rely upon my colleagues that work there to tell me what google does. As an operator of a network that may have more than one router on a lan, the reasons for this are generally one of two reasons: 1) Internet Exchange Point (we use BGP here to make forwarding decisions, not redirects) 2) HSRP/VRRP redundancy Even if you assume that someone picks the not-listed 3rd choice of placing their routers and some other (eg: dns, radius, dhcp) services all in the same subnet (poor network design, even for networking 101) making your fancy asic or processor based devices forward the packets correctly would seem to induce the proper situation. There is also nothing preventing these hosts (dns, radius, dhcp) from speaking some dynamic routing protocol to make better decisions. If there's more than one router on the lan, how do they communicate in your case? telekinesis? If these end hosts need to make "better" decisions, there is nothing preventing them from participating. The method you defend is redirects. The method I recommend is proper participation in the routing topology. > > Again, summary to the mailer: > > I still haven't seen any justification for changing the "MUST implement > Redirect functionality by a router" - there are sufficient use cases and > needs and also DoS issue with Redirect is weak reason to change the > MUST. Now, it's only SHOULD enable by default or not that needs debate. > For the 3rd time, since RFC 2461, router vendors enable Redirect by > default. Howsoever folks may thrash one cannot ignore legacy. So why > do we have to thrash so much and just agree to the SHOULD enable by > default? because my routers don't operate via telekinesis or redirects. my hosts don't listen to redirects as the information may be forged or improper. Your "haven't seen any justification" is purely smoke for making a host make routing decisions (via redirects; making it a router it seems?) but not requiring it to actually participate in the network. If you want your hosts to be routers, call them routers (and make them participate in the network) and don't try to play mix-and-match. - Jared -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
