Re: Minutes / Notes

[Iljitsch van Beijnum <iljitsch@muada.com>]

On vrijdag, jul 18, 2003, at 19:24 Europe/Amsterdam, Tony Li wrote:

> The problem is that we get to pick one.  We're making a fundamental
> change to the way the Internet works here.  There can't be two
> different architectures...
Now obviously if we're going to substantially change the way the 
internet works to accomodate multihoming, it is inconceivable that we 
leave out maintaining TCP connections through rehoming events.

On the other hand, if we can do multihoming _without_ substantial 
changes but this means breaking TCP sessions whenever a rehoming event 
occurs, that _could_ be a legitimate tradeoff.

But we haven't reached the point where we can make an informed decision 
on this yet. But my personal opinion on the matter is that "substantial 
change" is a red herring. The real question is how hard it is to 
implement a solution. The substantialness or fundamentalness of the 
change has little bearing on the implementation difficulty. The 
adoption of a round earth vs flat earth view is very fundamental but it 
doesn't change the way you navigate a sailboat through the 
Mediterranean Sea very much.

Iljitsch