[LINK] Net back-ups reviewed after cable breakdown

[Peter J. Cherny]

I said (inter -alia) to George Michaelson :-

>I'll stay off-list since this is probably a little to techie by my reading
>of the charter. (actually I'm not sure my original message was in-charter).

I'll take it now that others might be interested,
so a nearly acronym free summary follows ...

The failure of the SEA-ME-WE cable demonstrated that Telstra didn't have
a robust routing policy implemented for their Australian IP network.
The remedial actions taken to accomodate the cut severly disrupted the
network (and their web caches) for a day and continue to do so.

Telstra's practice has been to announce to the US backbone providers
that they have paths to all the networks in the ranges
203.0.0.0 - 203.34.255.255, 203.55.0.0 - 203.58.31.255 and
203.59.0.0 - 203.63.255.255 some of which are allocated by Telstra to it's
clients and others which are used by third parties who have connections 
to Telstra and (maybe) other networks.

There have always been a few exceptions, but on the whole the all addresses
known as 203.0.0.0/10 are announced to the rest of the world as a single
entity.

Telstra proposes to now announce each network as a separate entity
http://telstra.com.au/bigpond/direct/203-10-long.htm
creating an extra 1900 odd entries in the world's routing tables.

In a effort to keep the size of these tables within control, the backbone
networks only accept routes that are of a contain minimum number of addresses
usually what is known as a /19 or /20 (19 or 20 significant bits).

Unfortunately, most of the 1900 proposed 203 blocks are /24 (256 addresses)
and will be ignored by the major backbones unless Telstra makes specific
and cumbersome arangements.

In the short term, this is likely to isolate many of these networks from the
global backbone.

There are solutions that could be implemented at the interchange points
with the backbones (in the US), but I believe it would be much better
to ignore the knee-jerk reaction to this outage and engineer a robust
routing policy that allows Telstra to deliver a reliable service
regardless of any single-point-of-failure.

pjc

At 11:49 27/11/00 +1000, you wrote:
>> Yep, it's time for the technical press to not take things at 
>> face value ...
>
>Peter,
>
>Who said the SMH, the Australian, the Fin were *technical* press? Look up
>The Register for an explanation of the acronym "PFY"...
>
>Re the route notification you attached, the semi-technical (me, for eg)
>might like more info on the impact...
>
>Richard