> From: "Michel Py"
> I realize now that this is where we erred: by shifting the
> multi-homing problem from the ISP to the end-user, we made a
> less-palatable protocol ...
> We forgot to KISS.
I'm going to have to seriously disagree with the philosophical
principle(s) you seem to be promulgating here, because I think you're
Starting with the more specific principle, moving functionality from "the
network" to "the user entities" is not at all necessarily the Wrong
Thing. Sometimes it's very much the Right Thing.
A classic example is reliability - moving that from the network to the
user entities was definitely the right thing. For another example,
congestion control was also moved from the network to the end entities.
(I know there is still some doubt about the wisdom of this, but we have
made it work, and certainly nobody is talking seriously about moving it
In analyzing whether moving functionality out to the edges like that is
the right engineering decision, one has to look at a whole range of
factors, and look at the whole system - and sometimes it is indeed the
Right Thing to make the end entities more complex.
More generally, there's TANSTAAFL. We can't expect to build a global
system with all sort of interesting capabilities, and do so with the
same architecture we had when we were planning a much smaller network
with more limited capabilities.
There is going to have to be extra complexity to get those extra
capabilities, and inevitably (for good system architectural reasons, as
discussed above), some of that complexity is going to wind up in the end
Ietf mailing list