[prev in list] [next in list] [prev in thread] [next in thread] 

List:       zebra
Subject:    [zebra 293] Re: Linux BGP4 router
From:       Tony Li <tli () juniper ! net>
Date:       1999-02-07 23:40:01
[Download RAW message or body]


hgredler@xylan.com (Hannes Gredler) writes:

> > It's common practice. Saying it's "*not* permitted" is just silly.
> 
> let me be more precise; - doing so is technically permitted but from
> an internet architectural standpoint not desireable and as you mentioned,
> very nasty; - well using ipv4 this is common practise and works -
> but the internet is penetrated by small prefixes (/23, /24) and
> lots of as-paths flying around.
> surely not very scalable ....


True.  However, hole punching (more strictly - exporting more specifics) is
certainly expected in the case of multi-homed sites.


> well i'm not sure if this (hole punching) will possible in a next gen.
> internet (v6) which scalability depends on a global aggregation scheme.
> n.g. protocols like idrp are heavily dependend on multi level aggregation.
> but i assume in this new world everyone wants to be a "top level
> aggregator" as today every owner of an as number is :-)


Interestingly, the scalability of the routing system does not significantly
improve by the deployment of V6.  This is because the scalability of the
routing system is driven by the number of top level prefixes that are found
in the default-free zone of the Internet.  The length of the prefix is only
a small constant multiplier (which, of course, is better for V4 than V6).

What's MUCH more relevant is the address assignment architecture.  Any
architecture that leads to a linear growth in the number of top level
prefixes will generally tend to limit the growth of the Internet.  Thus, I
would not expect to see a lot of additional more specific prefixes in the
V6 routing system.

Tony

[prev in list] [next in list] [prev in thread] [next in thread]