Hyperbahn: Place an upper bound on ttl

[Raynos]

Hyperbahn currently forwards with unlimited ttl.

This is a resource leak; we should fix it in two ways.

• Apply a maxPending check on peers.
• Enforce that the ttl is no larger then MAXIMUM_TTL ( 30 seconds ? ).

[prashantv]

How should this interact with streaming?

[Raynos]

@prashantv

There has been consensus that the ttl for streaming is "time until first response frame".

For streaming to work you just have to send a streaming response within 30 seconds.

[prashantv]

I meant how streaming's lack of timeout can very easily cause resource leaks. It means that any bug in a streaming server/client will leak resources on Hyperbahn, which seems pretty dangerous.

[Raynos]

@prashantv

Hyperbahn will figure out how to avoid resource leaking once we implement fixed object pooling. At that point once we can't allocate any more "objects" we just return Busy or something.