Hi Mark, > OK, all I am trying to say is that mapping IPv4 into IPv6 > address space > (ala 6rd) lets you get away with less state and configuration > that would > be needed without such a mapping. Deployments which would need more > shared state and/or config (which is generally related to the > number of > endpoints) have more to gain than those that do not.
Understood, but minimal state can enable some compelling capabilities including PI, native IPv6 prefixes that do not embed an IPv4, MTU handling, etc. Thanks - Fred [email protected] > - Mark > > On 11/9/10 10:18 AM, Templin, Fred L wrote: > > Mark, > > > > > -----Original Message----- > > > From: Mark Townsley [mailto:[email protected]] > > > Sent: Monday, November 08, 2010 5:57 PM > > > To: Templin, Fred L > > > Cc: [email protected] > > > Subject: Re: [Softwires] Topology of the homenetwork in > > > draft-tsou-softwire-gwinit-6rd-01 > > > > > > On 11/9/10 9:08 AM, Templin, Fred L wrote: > > > > > After the 6, 7 or 8 figure range in terms of number of sites > > > > > 6rd is enabling, the advantages vs. stateful methods > > > > > become quite apparent. > > > > > > > > I'm not sure that is exactly true. If there is concern for > > > > scaling, simply add more routers/servers the same as for 6rd. > > > I'm referring to the overall configuration, operation and > > > combined state > > > of the tunnels themselves regardless of the number of > routers. Setting > > > up and maintaining hundreds or even thousands of > configured, stateful, > > > tunnels is very different than hundreds of thousands, > millions or tens > > > of millions. Alternatively, 6rd scales from small to large to very > > > large, at the cost of bits in the IPv6 address rather > than individual > > > configuration and state. How expensive those bits are (or > > > should be) is > > > the subject of debate in various RIR forums. If we take for > > > granted they > > > are not free, then that needs to be considered when deciding which > > > tunneling option to deploy. > > > > I'm not talking about configured tunnels; I'm talking about > > automatic ones with a point-to-multipoint model. And, I'm not > > talking about lots of state; I'm talking about minimal state > > (say O(10) - O(100) bytes) on a per-neighbor basis. That should > > provide fine scaling for a single router/server, and if further > > scaling is needed just add more router/servers. > > > > Thanks - Fred > > [email protected] > > > > > - Mark > > > > > > > Thanks - Fred > > > > [email protected] > > > > > > > > _______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
