Using Protocol::HTTP in a non-Async-http application

[Forthoney]

In my search for a good, pure-ruby HTTP Parser, I ran into Protocol::HTTP. I think it's perfect for my use case, but I'm having some trouble actually integrating it into my project. Pretty much all of the examples across the Protocol::HTTP libraries assume it's being used in the context of Async-http, making it difficult to figure out how things may work without Async-http.

I'm currently stuck here. My application currently hangs at L23. At this point, the response has already been properly formed, but it fails to actually deliver it to the client (i.e. my browser). When I turn the server off via SIGINT, the response is displayed.

The code is essentially identical to Async::HTTP::Protocol::HTTP1::Server because that's the only other use of this gem in this use-case I could find. My assumption is something I have in there doesn't work without Async, but I'm also not very familiar with the async gem to debug this.

• Note that the Adapter.initialize's signature has been changed to also take in a logger object. This was done to get around the global object sharing limitation of Ractors

[ioquatix]

This is a reasonable use case and it's great that the gem/code is useful to you.

I don't see anything wrong with your code, but read_request also shouldn't hang unless there is some problem with the request. Are you able to give me a simple repro script so I can run it on my end?

[ioquatix]

Okay, it sounds like I should add persistent connection handling to my example and maybe we'd find the issue?

[Forthoney]

Okay, it sounds like I should add persistent connection handling to my example and maybe we'd find the issue?

Yes this would be amazing. Thank you so much for the help!

[Forthoney]

@ioquatix Don't mean to nag you but let me know if there's any updates!

[ioquatix]

Can you check socketry/protocol-http1@4b46cff and let me know how you get on.

Make sure you read the request body if any before reading the next request.

[Forthoney]

Yes this works great. Thank you again for the help and the great library!

[Forthoney]

Hi @ioquatix , I think I may have found an issue with the server.rb example. It works fine with curl, but when testing the connection through a browser, it doesn't recognize when the browser disconnected. This is an edge case likely due to the browsers doing some magic behind the scenes, I think it's worth getting right considering many people like to test things out on their browser.
Here's a screen recording of what happens.

screencapture.mp4

[Forthoney]

Ahh I found the issue - it is indeed due to browser magic lol. Apparently browsers hang on to connections for a specific time period (~2 minutes for firefox) even after a window is closed. Apologies for the confusion!

[ioquatix]

I didn't know that either! Nice find. Actually, browsers can share HTTP/2 connections between tabs too IIRC.