[menog] Peering

[Mike Hughes]

Mike Hughes wrote:

> I'm not saying it's a silver bullet. NDIX worked in the end because 
> enough people wanted it to work. But it wasn't an easy ride.

Bad form to follow one's self up, but I seem to have not gotten my point 
across. I've been wrongly percieved as disagreeing with Bill, when my 
point was actually the opposite and I was trying to put a statistic on it.

So, I'll try again, and then leave you in peace.

Attempts to run interconnected/wide-area IXPs (in the true sense of an 
IXP) run into problems, and generally just don't work. There's plenty of 
failed attempts, of which Bill and Philip can tell the stories if you're 
ever interested.

Out of all the schemes I've ever heard of to build such a thing, I only 
know of *one* which managed, and only under some fairly specific 
circumstances (vast amounts of good-will and sheer determination by a 
number of people).

I'm frequently approached by people suggesting I extend the LINX to new 
and far-flung places. I've yet to be persuaded that any of these 
suggestions would achieve anything besides increased cost for the 
exchange participants.

It's also hard enough to build a business case to extend to a new co-lo 
within the same metropolitan area, and I've got recent experience of 
this. Never mind extending to a different country.

The fact that there isn't a LINX switch located outside of the London 
area should be enough to demonstrate that the odds are against it.

Simple is good, cheap, and people will use it.

Mike