Re: [TLS] [AVTCORE] WG last call of draft-ietf-avtcore-rfc5764-mux-fixes-05

[Ilari Liusvaara <[e-mail]>]

On Thu, Mar 03, 2016 at 09:48:00AM +1100, Martin Thomson wrote:
> 
> [3]  I actually hope that we can change DTLS 1.3 so that it won't mux
> properly.  That will have a size benefit that should outweigh the cost
> of having to rev 5764 for 1.3.

I thought about this a bit...

It occurs to me that a way to compress most headers to 2 bytes (from
present 5) is:

Unencrypted: 16 03 01 xx xx or 15 03 01 xx xx
Encrypted: 80|xx xx

Where xx xx is the length bytes.

The unencrypted headers need to be kept for backward compatiblity.

The ranges don't overlap, since the first xx can be at most 0x41
(which occurs for maximum 16640 byte encrypted record payload),
so the first byte would be 0x80-0xC1, which doesn't overlap with
0x15 and 0x16 (which are used for unencrypted packets).

That would interfere with muxing however (requring at least
changing how muxing is done, e.g. by inserting a marker byte
for DTLS packets).


-Ilari