You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We need WPT tests for sendOrder... but this is hard to do.
The big problem is that to actually test this we need to somehow block transfers temporarily, let data get queued on multiple streams, and then see what order data goes out in after the block is lifted. We might be able to somehow cause enough congestion to let us buffer, but that's hard. Better would be to somehow tell the server to stop reading data for a while, somehow.
The text was updated successfully, but these errors were encountered:
Just to expand on the difficulty here, Firefox's stack currently applies backpressure based on flow control, but not congestion. This is a point of contention for us, so we might resolve that, but it highlights potential problems. My preference is that we look at flow control as the basis for this testing, but that means acquiring better control over how the test server (aioquic) operates in some detail. Detail that might fail if they decide to evolve their product.
The big problem is that to actually test this we need to somehow block transfers temporarily, let data get queued on multiple streams, and then see what order data goes out in after the block is lifted.
Is the problem how to build up multiple competing queues to dequeue from/roundrobin on? Won't the following do that?
We need WPT tests for sendOrder... but this is hard to do.
The big problem is that to actually test this we need to somehow block transfers temporarily, let data get queued on multiple streams, and then see what order data goes out in after the block is lifted. We might be able to somehow cause enough congestion to let us buffer, but that's hard. Better would be to somehow tell the server to stop reading data for a while, somehow.
The text was updated successfully, but these errors were encountered: