[arin-ppml] Rationale for /22

[Owen DeLong]

On Aug 3, 2009, at 8:30 AM, Stephen Sprunk wrote:

> Michael K. Smith wrote:
>> What was the rationale for the /20 in the first place?  Was it more  
>> than an arbitrary number?
>
> IIRC, it was a compromise between the folks wanting to be PI and the
> folks who didn't want their routers to explode.  When routers got
> bigger, the minimum went down to /22.
>
>> I can't see any detraction from getting providers to get an ARIN- 
>> assigned /24 instead of having to get a /24 from one provider and  
>> route it out another, being historically on the "purchasing" side  
>> of that arrangement.
>
> There is one major difference: if you get a /24 from your upstream and
> other folks in the DFZ filter it, you can still be reached via your
> upstream's aggregate.  If you have a PI /24, there is a much greater
> chance of breakage.
>
Yes, and, no.  If you have a PA /24, then, it creates the illusion  
that the
behavior you describe above is somehow acceptable.  However, the
reality is that such filtration seriously jeopardizes the purpose of  
multi-homing
in the first place -- the ability to survive failure of one of your  
upstream
providers.

If you have a covering aggregate and the provider advertising the  
aggregate
is the one that fails, filtration of the other more specific creates  
just as much
breakage, but, has the additional downside that said breakage could be
much more transient in nature and much harder to identify and/or  
troubleshoot.

Nonetheless, the consequence to the routing table is identical.

Owen