Re: [Technical Errata Reported] RFC7230 (4667)

[Willy Tarreau <[e-mail]>]

On Thu, Apr 14, 2016 at 12:20:12PM -0600, Roy T. Fielding wrote:
(...)

Thanks for the detailed history Roy!

(...)
> Apache httpd's chunk parser doesn't make use of chunk-ext unless it is bypassed
> (i.e., a module can implement its own chunk parser).  The core will allow up to
> 10 bad whitespace between the chunk-size and chunk-ext (to allow for space-padding
> of the chunk-size in fixed buffers), and then skip anything other than invalid
> control chars between the ";" and the end of line; i.e.,
> 
>    chunk-ext      = 0*10<BWS> ";" *( OWS / VCHAR / )
> 
> IOW, that parser would support such a fix, but only by accident.

In haproxy it's basically the same. It doesn't make any use of these
extensions but happily skips these spaces if they are there.

> I seriously doubt
> we can assume that arbitrary HTTP implementations would expect whitespace in those
> places.  It was a stretch to expect that all implementations can handle chunk-ext,
> but at least for that we had examples of how they would be implemented.

At least now we have an indication that at least one user exists :-)

> I don't have a problem with adding whitespace back in there, but I am not at all
> confidant that such a choice would be less likely to break things.  I don't want
> to play errata ping pong.

Well after all that's pretty much what the BWS was for : "you may have to
accept it for improved backwards compatibility but you must not emit it".

Regards,
Willy