[ppml] Multihome Pro Con Document

[Howard, W. Lee]

> -----Original Message-----
> From: Randy Bush [mailto:randy at psg.com] 
> Sent: Sunday, October 15, 2006 10:26 AM
> To: Howard, W. Lee
> Cc: ppml at arin.net
> Subject: Re: [ppml] Multihome Pro Con Document
> 
> > End-user networks may require some means of switching from two
> > types of residential access, but may be generally assumed to have
> > no more than two simultaneous links, a single prefix, and no need
> > for traffic engineering.
> 
> why the assumption of only two and why the assumption of no te?
> i want voip out the left and quake out the right, for example.

I use adverbs carefully.  

You're special, and clueful, and can build complex networks.  
If we're not allowed to make assumptions, we can't define requirements,
and can't do anything.

> > Enterprise networks need multihoming primarily for reliability,
> > but may also need some level of best-path selection, and may need
> > some degree of load distribution.  In IPv6, they will generally
> > have no more than two prefixes.
> 
> wait a sec!  i was told that we would not have the prefix explosion
> in v6 because we could give each site enough space so that it would
> only need to announce one prefix into the global routing state.

Did you get a receipt?

Most enterprise networks should only have one prefix, but some may
need multiple /56 (depending on ISP policy) or even /48.

If we can converge on 1-3 technologies, then they can be applied 
where needed.  A residential customer may desire an ISP technology,
but finding a provider who can support it may be tricky.

Lee

> 
> randy
>