Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

Loading…

More efficient share connections with unconsumed bodies #123

Closed
snoyberg opened this Issue · 5 comments

3 participants

Michael Snoyman Felipe Lessa Greg Weber
Michael Snoyman
Owner

Should Warp determine whether or not to close a connection based on the size of the remaining request body?

Consider the case where the a client wants to send two requests to the server. The first request has a 2MB request body, which the application entirely ignores. Warp currently will flush that request body and then process the next request.

Perhaps instead, Warp should analyze the remaining request body size, notice that it's over a certain threshold (to be determined via benchmarking), and instead close that connection. The theory is that sending that unnecessary 2MB body over the network presents a higher overhead than creating a new connection.

Felipe Lessa
Owner

Is there a precedent? Won't clients think that there was an error?

Michael Snoyman
Owner

Firstly... how did you see this so quickly? Is there some Github setting to get emails for all organization issues?

I don't know if there's precedent, but I don't think it will cause a problem. We're going to send a complete response to the client before closing the connection, which should let it realize that everything went OK. But it's certainly something we need to check first.

Felipe Lessa
Owner

http://25.media.tumblr.com/tumblr_mcglkogOk31qh1l5wo2_400.jpg ;-)

Yes, I receive an e-mail for each and every issue created on my watched repos, and by default I watch all repos from my organizations (I may opt out, though). However, I do not watch all threads by default, so I wouldn't receive a message for your reply if I hadn't replied or explicitly watched this thread first.

Greg Weber
Owner

The Watch button at the top will change your settings.

Michael Snoyman
Owner

Implementing this will end up raising even more questions. For the time being, let's just drop this idea.

Michael Snoyman snoyberg closed this
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.