Matt Caswell

Note this is very much a straw man...

>> An application can (optionally) insert new connections into the connection queue directly by calling SSL_EVENT_CTX_add_new_connection(). This would also result in a "new connection" event being issued. > > Ordinarily...

> I think there's too much semantic confusion here. The SSL object representing a default stream is being used to represent two things: the connection, and a stream. To make...

> I have only glanced over the recent updates/discussion, but I recommend not having a notion of "default" stream and completely separating out the connection and stream objects. It's entirely...

> Pretty much ever QUIC API I know of is either a completely separate, new API or a new combined API was written on top of something like TAPS. While...

> This means that the number of sockets, and their lifespan, is dependent on complex internal details of QUIC, and asking an application to figure this out is unreasonable. I'm...

> Are we assuming connection migration is always initiated by the application? Generally I would expect connection migration to be triggered by OS-level networking configuration changes. This was my assumption,...

> These decisions as such don't necessarily have to be made today, but the API decisions they relate to do. As I see it, it would be best to hedge...

>> Are we assuming connection migration is always initiated by the application? Generally I would expect connection migration to be triggered by OS-level networking configuration changes. > > This was...

I've pushed a number of tweaks and updates to respond to the feedback so far.