Re: Sending Referer [#144]

On Mon, Jun 1, 2009 at 12:14 PM, Brian Smith <brian@briansmith.org> wrote:
> There's no way to represent the null value in a way compatible with RFC 2616
> syntax.

Why is that?  The RFC 2616 syntax appears to be happy with an
absoluteURI.  We need only represent the null value using an
absoluteURI.

> New syntax for existing headers should be done only for HTTP 1.2 or
> later. Further, this would be a new feature, which would make it out of
> scope for HTTPbis.

This appears to be clarifying the semantics of the Referer header, not
adding a new feature.  As in "if you see a Referer header with this
value, then that means there wasn't a referring URI or the user agent
got confused and couldn't figure out the referring URI properly."

> I think that the less that is said about Referer in the HTTPbis
> specification, the better. Its syntax and intended purpose is enough. Any
> requirements regarding Referer for UAs are pointless; they will be ignored
> by most UAs, especially UAs that aren't browsers. A specification for what
> web browsers should do with the Referer should go into a browser-specific
> specification.

As things stand, the document forbids user agents from always sending
the Referer header, preventing a browser-specific specification from
requiring this behavior.

Adam