On Fri, Dec 07, 2018 at 05:48:40PM -0500, Jared Mauch wrote: > > On Dec 7, 2018, at 5:44 PM, Nico Williams <nico@xxxxxxxxxxxxxxxx> wrote: > > On Fri, Dec 07, 2018 at 03:46:15PM -0500, Jared Mauch wrote: > >> 1) We have long-lived TCP sessions, measured in years. (Implied: many > >> of the transport people really prefer stable routes without > >> flapping/jitter/reordering from us) > > > > <rhetorical> > > I've long wondered why BGP has to be this way and why it still is after > > so many years. > > </rhetorical> > > I have seen BGP sessions up for over 7 years on some routers. One can > argue if this is a good thing or bad. It’s quite an interesting thing > to see and go “wow, someone forgot about this one …” What I meant is that I don't understand why resetting a connection should still cause routes to flap. Instead I'd expect session management negotiation to determine whether the thing that failed (e.g., a BGP daemon) implies that the routes are gone or not. Then RST injection would not be a problem. You'd still need integrity protection, but then that could always have been done with TLS. Basically, it feels like BGP is stuck in the stone age. I'm not asking the routing area to fix it. I'm not saying that we should not cater to their needs. My point truly was rhetorical. I do hope the routing area does improve session management in BGP some day. Nico --
