Re: An alterntive approach regarding i93: Repeating Single-value headers

On ons, 2008-01-02 at 19:37 -0800, David Morris wrote:

> In my mind, this means we can ignore cookies, if we choose, because this
> revised wording makes it an error to fold an unknown header.

Which is a big change. The current wording always allows combining of
any repeated header, wellknown or unknown extension header never heard
of before, with Cookie being the single (unspecified) exception.

If combining repeated headers renders the message invalid then it was
invalid before the headers was combined. I.e repeated sinngle-valued
header. The only issue (apart from Cookie) is that combining headers may
actually render invalid messages valid by removing the duplicated header
(now a single header with a longer value)

> It also
> allows flexibility with future extensions by allowing the designers to
> include repeated headers if that makes sense.

A repeated header is by definition a list header, and is the default
header format unless known otherwise.

It's better to mention Cookie as an exception, with an informal
reference to the cookie rfc..

Regards
Henrik