PSA: Removing support for DTLS-to-SDES fallback

621 views
Skip to first unread message

Taylor Brandstetter

unread,
Sep 22, 2017, 2:13:53 AM9/22/17
to discuss-webrtc
If you don't use SDES, you can stop reading.

With this CL, we're removing support for falling back from DTLS to SDES. In other words, applying an offer with both "a=crypto" and "a=fingerprint", with an answer that chooses one or the other.

The reasons for doing this are:
  • It's non-standard (and in fact explicitly goes against at least a couple "MUST"s).
  • It makes our code more complex.
  • It's only even possible via SDP munging, which will eventually be illegal anyway.
  • It moves us one more step towards removing support for SDES, which we have to do anyway (it's a "MUST NOT" in JSEP).
If you have a compelling reason why we should continue supporting this, please speak up now before this change makes it into a Chrome release branch. Thanks!

Urs D.

unread,
Oct 2, 2017, 10:56:11 AM10/2/17
to discuss-webrtc
Hi Taylor,

so doing SDES rightaway (not falling back from DTLS) will still be supported for now?

Taylor Brandstetter

unread,
Oct 2, 2017, 8:13:38 PM10/2/17
to discuss-webrtc
Correct. Though, as mentioned, this is a "MUST NOT" in JSEP. So we'll need to remove support for it eventually to be standards-compliant.

--

---
You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss-webrtc+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/discuss-webrtc/0bc14e17-2696-47a7-8142-15c076ca6cbc%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

Andy Lee

unread,
Oct 5, 2017, 8:29:16 AM10/5/17
to discuss-webrtc
Which Chrome release will this change be seen?

Andy


On Monday, 2 October 2017 20:13:38 UTC-4, Taylor Brandstetter wrote:
Correct. Though, as mentioned, this is a "MUST NOT" in JSEP. So we'll need to remove support for it eventually to be standards-compliant.
On Mon, Oct 2, 2017 at 7:56 AM, Urs D. <ursdeu...@gmail.com> wrote:
Hi Taylor,

so doing SDES rightaway (not falling back from DTLS) will still be supported for now?


On Friday, September 22, 2017 at 8:13:53 AM UTC+2, Taylor Brandstetter wrote:
If you don't use SDES, you can stop reading.

With this CL, we're removing support for falling back from DTLS to SDES. In other words, applying an offer with both "a=crypto" and "a=fingerprint", with an answer that chooses one or the other.

The reasons for doing this are:
  • It's non-standard (and in fact explicitly goes against at least a couple "MUST"s).
  • It makes our code more complex.
  • It's only even possible via SDP munging, which will eventually be illegal anyway.
  • It moves us one more step towards removing support for SDES, which we have to do anyway (it's a "MUST NOT" in JSEP).
If you have a compelling reason why we should continue supporting this, please speak up now before this change makes it into a Chrome release branch. Thanks!

--

---
You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss-webrt...@googlegroups.com.

Taylor Brandstetter

unread,
Oct 5, 2017, 12:39:43 PM10/5/17
to discuss-webrtc
M63. You can test right now with the Canary or Dev channels.

To unsubscribe from this group and stop receiving emails from it, send an email to discuss-webrtc+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/discuss-webrtc/9a5cba6f-07a9-46c0-a4f8-d3d45c127b5b%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages