OT:Internet performance ...

[The Natural Philosopher]

There is, more or less.

[The Natural Philosopher]

yes, that is true. But it doesn't help..

You could also construct a malware or media harvesting system using exactly the same tools.

In essence, you want to have an identity for legitimate stateful transactions, but not have one for illegitimate transactions that you dint authorise. .

Cake and eating it comes to mind.

[Tim Streater]

May depend if the routing updates are over a congested link.

[Tim Streater]

Now you're being picky.

Before I retired, I was involved with a large European-wide backbone, connected at several points to a couple of worldwide ISPs (connections were typically at 10Gbps). Each of those connection-point routers had two, three, or more copies of those ISPs' view of the Internet, typically containing 200,000 routes each. Prolly more now as that was more than three years ago.

Plenty of routes to update if something happens - in particular if someone's parameters somewhere are not tuned very well and you find you're receiving 50k updates a second. Flapping routes can make it even worse.

[The Natural Philosopher]

No. it doesn't. If there is no response after (two ?) minutes from a BGP peer for ANY reason, the router will mark the link down and look for an alternative strategy, and propagate the fact that its down to the 'global routing tables' as one poster put it.

[The Natural Philosopher]

yes. There is no one single global routing table at any point in time, but there is certainly something resembling it all over the internet: where you are determines what you use of course - a big ISP will aggregate its internal routes into a broader statement to advertise..there is no need for the outside world to know the intimate details of how to reach my PC, all they have to know is 'give that packet to IDnet and they will sort it out themselves', since they are the last 'border' on the network leading to it.

HOWEVER in terms of BGP there exists at any given time a dynamic table that does reflect global routing between ALL entities who use BGP at their edges.

Its like saying 'there is no Usenet database or server' Well maybe no, not one, but the data in Usenet should all be the same on however many server it has storing it. Plus minus a few minutes.

[Tim Streater]

The router to which the flaky link is connected, yes. But I'm talking about distant routers on other networks receiving updates for their routing tables. Those updates could come via congested links.

[The Natural Philosopher]

The internet core is not large in tersm of nodes. and a congested link won't take minutes to exchange a few BGP packets.

[dennis]

At any one time there are several "global routing tables". Updates travel asynchronously so the tables held by the routers don't match all the time.