Re: draft-ietf-httpbis-replay-03, "5.1. The Early-Data Header Field"

> On 15 May 2018, at 7:28 pm, Willy Tarreau <> wrote:
> FWIW I've used many times "Host:" with no name when connecting to an
> HTTP/1.1 server without knowing its name, and rarely faced any issue
> with this. I'm not saying it's right or wrong, but that many
> implementations support an empty header field and distinguish it from
> an absent one.

Sure - although that's pretty broken for Host, this isn't a completely uncommon thing. When we did SOAPAction way back when (shudder), we came to the conclusion that some HTTP implementations might drop empty headers, so it wasn't safe -- but I don't remember how much data that was based upon.

That said, just because it's possible with HTTP headers doesn't mean it needs to be possible with structured headers.


Mark Nottingham

Received on Tuesday, 15 May 2018 09:38:40 UTC