[AusNOG] Small Pipe BNE/Agile issue

[Matthew Moyle-Croft]

James Spenceley wrote:
> Yes, I do and the great thing about BGP is there are so many tools.
So many square pegs, round holes and idiots with hammers :-)
>
> The argument that I should have the *right* to decide where my traffic 
> goes, is in no way more valid than I should have the *right* to decide 
> how I get my traffic. In the end its a technical function in a router 
> ... as path beats neighbour address, lpref beats AS path, specific 
> route beats lpref. Everyone has the required tools to implement their 
> policies.
The problem with this approach is that your policy isn't more or less 
valid than mine.   DJH pointed that out previously.   So, we come back 
to - there are many tools, but just like using a lawnmower before 7am on 
a Sunday morning, using them to assert that your policy is more 
important than mine can be considered "not nice" in a shared environment 
like an MLPA IX.
>>> I'm not asking anyone to accept >/24s just saying that it actually a 
>>> very valid way to traffic engineer, some people seem to have take 
>>> offense at being sent routes >/24 and if they want to filter go 
>>> ahead. I'll keep my options open.
>> So, would you like a fence to sit on here James? :-)
>
> Ha, fair cop. So here it is in black and white. A lot of the world 
> relies on AS path for path selection, or incorrectly uses local-pref. 
> I think its valid that we de-aggregate to the Peering / MLPAs to 
> ensure the most amount of traffic stays on the IX. If you have a bunch 
> of traffic rich /24 then there is no reason you shouldn't advertise 
> the /25 to get the same effect.
My experience is that deaggregation often leads to more rather than less 
TE issues.   Especially when you have transit providers who are IX 
members - which is what started this whole debate when Skeeve complained 
that we were passing on a /26 from AS4802 which meant he saw traffic via 
transit (AS4739) rather than PIPE IX in BNE.  

This lead to my whole issue with people deaggregating beyond /24s at 
PIPE and the lack of control with MLPAs in general.

So, maybe, we should continue this debate at AUSNOG02?

MMC