Hello!

On Mon, Mar 05, 2018 at 11:52:57PM +0800, Haitao Lv wrote:

[...]

> > Overall, the patch looks like a hack and introduces too much
> > complexity for this feature. While I understand the reasoning,
> > the proposed implementation cannot be accepted.
>
> Could you clarify that whether is this feature not accepted or this patch?
>
> If this feature is not needed, I will terminate this thread.
>
> If this patch only looks like a hack, would you like offer any advice to write
> code with good smell?

We've previously discussed this with Valentin, and our position is
as follows:

- The feature itself (autodetection between HTTP/2 and HTTP/1.x
protocols) might be usable, and we can consider adding it if
there will be a good and simple enough patch. (Moreover, we
think that this probably should be the default if "listen ...
http2" is configured - that is, no "http1" option.)

- The patch suggested certainly doesn't meet the above criteria,
and it does not look like it can be fixed.

We don't know if a good and simple enough implementation is at all
possible though. One of the possible approaches was already
proposed by Valentin (detect HTTP/2 or HTTP/1.x before starting
processing, may be similar to how we handle http-to-https
requests), but it's now immediately clear if it will work or not.
Sorry, but please don't expect any of us to provide further
guidance.

--
Maxim Dounin
http://mdounin.ru/
_______________________________________________
nginx-devel mailing list
nginx-devel@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx-devel