Re: [rrg] IPv4 & IPv6 routing scaling problems

[Tom Vest <[e-mail]>]

On Feb 12, 2010, at 12:15 PM, Christopher Morrow wrote:

> On Fri, Feb 12, 2010 at 11:51 AM, Tom Vest <[e-mail]> wrote:
>> 
>> On Feb 12, 2010, at 11:45 AM, Christopher Morrow wrote:
>> 
>>> On Fri, Feb 12, 2010 at 11:41 AM, Danny McPherson <[e-mail]> wrote:
>>>> 
>>>> On Feb 12, 2010, at 9:39 AM, Christopher Morrow wrote:
>>>> 
>>>>> I really think that the conversation about ipv4/ipv6 and route-scaling
>>>>> has to understand that for the foreseeable future we're going to have
>>>>> to deal with both ip protocols... and in 25-30 (maybe more) years a
>>>>> third protocol.
>>>> 
>>>> Indeed, hence my "long term transitional coexistence" phrasing :-)
>>> 
>>> Sorry, I meant 'there is no transition, there is only coexistence'
>>> (from my perspective at least that seems to be what'll happen, of
>>> course no crystal balls and only 5 computers ever will be needed.)
>>> 
>>> -Chris
>>> (and I get that you == danny get this, but for the record I think we
>>> should be clear that ipv4 ain't going away, ever)
>> 
>> Does "ain't going away" mean that you cannot envision any time in the future in which, say, IPv4 has the same impact on scalability concerns as RFC1918 has today?
> 
> 1918 does have scaling implications, if you carry it as internal
> routes... at some ISP's internal routes are ~30% of the total table
> size seen. It's not the same cost for everyone, but neither is a
> single prefix in the global table :(

That had not escaped me ;-)

But it seems to me that those RFC 1918 implications are sufficiently different in degree (e.g., of extra-local effects, of susceptibility to effective local mitigation) that the question made sense -- i.e., that the comparison was more realistic than either of the alternative scenarios (A: IPv4-specific scaling impacts are never going to get any better, B: IPv4-specific scaling issues are certain to be completely eliminated).

> I also suspect (as I said before) we'll see longer than /24 routes in
> the global table (not leaks/mistakes) before this is all over. I don't
> like that concept, but I don't see an outcome that doesn't include
> this.
> 
> -chris

Well, you have company there.
Hopefully both of us will turn out to be wrong, or alternately that this development won't eliminate all detours around A.

TV