Re: [bfcpbis] Updates to draft-ietf-bfcpbis-rfc4583bis required to enable ICE

[Roman Shpount <[e-mail]>]

I would like to make to additional corrections to my proposed text.

1. The text for seciton 3 should be update to mention that BFCP version for
unreliable transports should be used in case of TCP/DTLS/BFCP:

TCP/DTLS/BFCP, which is realized by running BFCP *for unreliable transports*
on top of DTLS as described in this specification and running DTLS on top
of TCP is realized using the framing method defined in RFC4571, with DTLS
packets being sent and received instead of RTP/RTCP packets using the shim
defined in RFC4571 so that length field defined in RFC4571 precedes each
DTLS message.


2. In ICE considerations, I would like to add:

Using ICE with protocols other then UDP/TLS/BFCP and TCP/DTLS/BFCP is
outside of scope for this specification.


Thank You,

_____________
Roman Shpount

On Thu, Mar 2, 2017 at 6:49 PM, Roman Shpount <[e-mail]>
wrote:

> Charles,
>
> On Thu, Mar 2, 2017 at 6:29 PM, Charles Eckel (eckelcu) <[e-mail]
> > wrote:
>
>> [cue] We define the proto field value UDP/TLS/BFCP in this draft for BFCP
>> over DTLS. Would it not be more straightforward and consistent to define
>> the new proto value as TCP/UDP/TLS/BFCP instead of TCP/DTLS/BFCP?
>>
>>
>>
> I am trying to keep proto names as close as possible to
> draft-ietf-mmusic-dtls-sdp. I understand that there are already
> implementations which use UDP/TLS/BFCP so we cannot change it to the
> technically correct value which is UDP/DTLS/BFCP. After all, we are using
> DTLS transport, which is different from TLS.
>
> Since there are no implementations of TCP/DTLS/BFCP, we should use the
> technically correct protocol string. There is no UDP layer in TCP/DTLS/BFCP
> transport stack, since DTLS packets are passed directly to RFC4571 shim.
> Because of this I think TCP/DTLS/BFCP value is accurate and appropriate.
>
> Regards,
> _____________
> Roman Shpount
>
>