Re: [quicwg/base-drafts] DATA frame encoding is inefficient for long dynamically generated bodies (#1885)

[Mike Bishop <[e-mail]>]

I'm sympathetic to the desire to reduce framing overhead, and having previously described the frames as "regions," I'm not philosophically opposed to having a region run to the end of the stream.  However, we've built / inherited from H2 a model in which frames have a known length from the start (though some might be so large that you have to process them in a streaming fashion).  I'm cautious about up-ending that model and saying that some frames have indeterminate length, and you'll know the end when you get there.

Depending on your implementation model, that might be fine.  If you have a processor that gets fed the bytes until the end of the frame, then it already handles a stream and being told that the payload has ended after/as the last byte is written to it.  This is a scoped change to the layer that reads the bytes from the stream and stuffs them into the processor.  If you have a processor that gets told the length and consumes that many bytes from the stream, this might be more of an architectural change.

I'm curious to hear from implementers how much of a challenge they foresee this being in their architectures.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/1885#issuecomment-431447381