North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: More long AS-sets announced

  • From: Michael.Dillon
  • Date: Wed Jun 22 05:36:37 2005

> Many of us in the operational community are required to conduct testing 
> in lab environments, followed by well-announced maintenance windows. 

Thanks for this funny post. I needed a good laugh.

It has been years since people have needed a reminder that as the
biggest and most complex telecommunications network in the world,
the Internet cannot be tested in a lab because it is not possible
to construct a lab environment that matches the scale and complexity
of the Internet.

> Why is this operational test supposed to be given freer reign on the 
> 'net than our own operations?

It's not. Your operations are also an EXPERIMENT as are the operations
of every other ISP. People can pretend that this is not so and make
claims to the contrary in their marketing literature, but the Internet
by its very nature is and will always remain, AN EXPERIMENT. Your network
probably poses more threat to the net than the long AS experiment. That
is because you haven't tested all the possible configurations of 
mistakes in your announcements and therefore your peers do not know if
their routers can handle it when your network goes nuts. And no matter
how much you test it and how tightly you manage it, you can never get
that probability down to zero. At least the long-AS folks are warning
in advance that their area of the net could be acting strange soon. How
many other ISPs provide their peers with warnings about misconfigurations?

If, after this advance warning, your network falls over because of the
long AS tests then I suggest that your own lab testing has been 

--Michael Dillon