interledgerjs
Don't expire other connectors' routes
As far as I can tell, all connectors effectively need to have the same route broadcast interval and expiration time. If a connector has rates that don't change very often, they might not want to waste the effort rebroadcasting the same data over and over again. However, right now the other connectors would expire the routes. This may even create an incentive to have a lower than average time in case there are some connectors with lower times.
Routes should probably carry something like cache control headers so you know how long another connectors' routes are valid for, rather than expiring everything on your own interval.
It's important that its routes expire quickly when a connector goes down, but you're right that it makes no sense to retransmit them. So we think that we should have each connector send their routing table and then just a heartbeat + any changes to the routes.
That sounds good. We should also have a way to request the full table when you've gone down or if you think you've missed updates
On Fri, Dec 16, 2016, 6:29 PM Stefan Thomas [email protected] wrote:
It's important that its routes expire quickly when a connector goes down, but you're right that it makes no sense to retransmit them. So we think that we should have each connector send their routing table and then just a heartbeat + any changes to the routes.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/interledgerjs/ilp-connector/issues/277#issuecomment-267619164, or mute the thread https://github.com/notifications/unsubscribe-auth/ADHIkuuxh231q0wWXpgQFUkb8hWk_wGkks5rIq50gaJpZM4LOPno .