On 27/03/2017 14:19, Sergio Garcia Murillo wrote:
Hi Boris,
Re-reading the draft-03 the X bit is handling on the receiving side:
Insert all header extensions up to the OHB extension, but
exclude the OHB and any header extensions that follow the OHB.
If there are no extensions remaining, then the X bit MUST bet
set to 0. If there are extensions remaining, then the
remaining extensions MUST be padded to the first 32-bit
boundary and the overall length of the header extensions
adjusted accordingly.
So there is no need to signal the original X bit.
I think that we need to signal at least the padding bit, as if there is
any padding on the original rtp packet, it will be converted into
payload for the double encrypted packet, which would not carry padding
by default (we could force to keep the padding bit as the original
packet and set the padding count to 0 but that will add one extra byte
anyway).
Re: padding only bit, I don't have a strong opinion. But the only thing
is that currently padding on the original packet can only be of 256, so
"padded only" encrypted packet can't be bigger than that too.