Re: Header addition with HTTP 2.0

[Amos Jeffries <[e-mail]>]

On 1/04/2015 11:19 a.m., Martin Thomson wrote:
> On 31 March 2015 at 13:17, Bence Béky wrote:
>> Sorry, I forgot about that.  In that case, you are right, there is no
>> way to avoid re-encoding the header.
> 
> There is: if you aren't doing anything more than adding a header
> field, as long as you choose a literal representation and you also
> pass through all messages that relate to the header table (max size,
> everything) verbatim, you can shim this in.  You won't learn much in
> the process though unless you maintain your own table because you
> won't be able to decode header fields without the header table.
> 

And to carify for the record by "shim it in" you mean append/insert the
new header to the existing HEADERS payload using a "Literal Header Field
without Indexing" and incrementing the HEADERS length field to account
for the new bytes, right?

With of course the complex logical adjustments for HEADERS padding,
frame max-length and CONTINUATION frames.


PS. "Literal Header Field never Indexed" is also an option, but
restricting what recipient proxies can do to relay the HEADERS is not
always desired. Either of the not-Indexed encodings will work.

Amos