On Sun, Aug 8, 2021 at 2:47 AM Robert Raszuk <robert@xxxxxxxxxx> wrote:
> if someone uses anycast to establish long lasting TCP sessions (say 12h) - Good Luck !
Transport sessions cannot reliably last for 12 hours on the Internet.
That's the thing.
The thing is that 1000s of people come to their desks at 8:00, they turn on their financial application - which uses TCP - and turn it off at 17:00 local time. And the single session stays up just fine for 9h. In fact session stays much longer such that in such application there is hard stop at 18 hours to kill the session. And such forced session killing does happen a lot every day.
Of course - to make it crystal clear I would never endorse designing any application like this - but when such apps were written X years ago networking was taken as balck box abstraction shielded by a socket. It is just real proof that your point is simply not true.
> A single BGP route announcement takes some 30 seconds to propagate, and
> sometimes a route withdrawal takes more than 4 minutes.
That's a pure occurance of spreading FUD.
I do recommend running your own measurements or consult those who do it well. For example Geoff. In fact his cats and data presented illustrate very well the point I was making earlier:
And with the average of 4.3 of average AS path length maybe was something in late 1990s where withdraws or first updates were subject to MRAI timer in some implementations - that is just history.
Best,
R.
R.
