[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: v6 multihoming and route filters

On 6-jul-2006, at 14:36, bmanning@vacation.karoshi.com wrote:


What fallicy?



	where is the "global routing table"?  How is it defined?
Either as a particular instance (i.e., taken from one router) of the set of routes that allow connectivity to the internet at large or the set of prefixes that make up the address space used on the internet. Obviously this can be made as complex as desired. 


	In zero cases that I am aware of, can any given node running
	IP reach any other arbitrary node running IP.


What does that prove?


	I posit that this "global routing table" is a myth.  The -ONLY-
	thing that matters to me are are the entries in -MY- routers memory.
	How they get there is -MY- choice and I get to dictate what is placed
	there.  What you have in -YOUR- routers is your business.
That's nice for you but I'm glad that you're not my ISP because I don't think personal preference has an important place in running a global network. Not only because different people have different preferences, but also because it doesn't scale. 


But we were talking about inter-domain routing. Filtering on prefix
size is a very useful technique that does indeed have unfortunate
side effects. But if you want to get rid of this technique you'll
have to come up with something to replace it... Not filtering is not
a reasonable long-term strategy.
I never said not filtering was an option. Proxy aggregation is viable 
	as well.
Not with BGP as it exists today, and security issues are quite problematic with this. 


	IF this draft is to have any credibility, the only choice will be to
describe what prefix size filtering is, what the ramifications are, and what the "unfortunate side effects" are. As an ION, one might indicate 
	that given the implementation constraints of 2006, an operator might
	consider the merits of selecting certain prefix sizes, if they choose
to use prefix-sized filters in their routers. Two distinct documents.
I disagree. Having a prefix that works in 98% of all cases is not nearly as useful as one that works in 99.98% of all cases. Having a free-for-all with regard to prefix filtering is suboptimal. And I think now we're back to the original issue so this is a nice point to have a go at letting the discussion rest. :-) 


	IMHO of course.  The IETF will do what/as it pleases.


Given rough consensus, yes.