Re: [narten@us.ibm.com: PI addressing in IPv6 advances in ARIN]

[Marshall Eubanks <tme@multicasttech.com>]

On Apr 14, 2006, at 3:34 AM, Iljitsch van Beijnum wrote:

> On 14-apr-2006, at 2:05, Jason Schiller (schiller@uu.net) wrote:
>
> > I'd like to point out that if people think that PI (and placing 1 or
> > more routes in the global Internet routing table) is an acceptable
> > solution to multi-homing, then it is also likely that more  
> > specifics of PA
> > space to support multi-homing is also acceptable, and likely to  
> > happen.
>
> > This removes the fact that few people would be able to multi-home  
> > in the
> > traditional sense.
>
> Actually it doesn't. It's very hard to qualify for a /22 so  
> effectively this only helps Big Content. The issue

No, it is not. (Speaking as someone who has helped people get one.)

Regards
Marshall

> of smaller ISPs or whole sale transit carriers that don't give out  
> address space to their customers because those all have PA isn't  
> addressed here.
>
> But it completely changes the game: why spend time on deploying  
> shim6 if you can round up your IPv4 needs to get to a /22 or lobby  
> for reducing the /22 requirement?
>
> > So there are only two questions:
>
> > 1. Is there sufficent need to continue shim6 for a no-frills easy
> > multi-homing solution for consumer and small business customers  
> > that can't
> > or won't support BGP?
>
> > 2. Are people concerned with the long term implacations of de- 
> > aggrgegation
> > for multi-homing, and if so, then should we pursue a protocol  
> > solution
> > that has full support for TE in an effort to replace de- 
> > aggrgegation as a
> > solution and halt (or possibly reverse) the growth of the routing  
> > table?
>
> Let me add a third question: will shim6 be deployable when people  
> can get PI?
>
> Maybe shim6 can still do some good but we really need to rethink  
> our goals here. I was uncomfortable with moving fast and only  
> implementing bare bones mechanisms before, and even more so now. We  
> should slow down and get it right the first try because if there is  
> a place for shim6 in a world with IPv6 PI, that place is for a GOOD  
> shim6 solution that solves the problem well, something that kinda  
> works just to get something out the door isn't good enough and  
> there won't be a second chance to come up with an improved version.
>
> > BTW, I was one of six people who voted against moving forward with  
> > PI.
>
> How many people voted in favor?