Thank you for the explanation. I definitely agree that this sounds very
important. My concern remains that the parameter name and
developer documentation does not convey this and without this context it's
unclear why I should use this. As an example, the config parameter could
just be called negotiateDataChannels, as a cue that this will ensure your
connection is configured to support creating data channels. It would be
worth also explaining why you want to do this, with a short snippet showing
how a data channel can be added later.

The spec says that it is negotiated before audio or video but doesn't
explain why you might want to do this - i.e. to ensure that you can
negotiate a connection even if the audio or video channels fail to
negotiate a compatible codec.

TLDR; I'm not opposed to this feature - after learning what it does and
why, I agree we need it. I just think it would be very hard for someone
reading about this to understand that they should use it to resolve these
issues.



On Mon, Jan 26, 2026 at 7:31 AM Henrik Boström <[email protected]> wrote:

> This simple API plugs a missing piece of functionality: there is no way to
> tell WebRTC that you may want to use data channels at some point in the
> future without having to re-negotiate, so the only way to ensure you
> negotiate data channels unconditionally is to create these dummy data
> channels not to be intended for use that you have to manually ignore at the
> other endpoint, both with regards to events firing and when parsing
> statistics.
>
> Creating dummy data channels is a workaround to a mistake in the API in my
> opinion, and I would support it even if all that it achieved was API
> ergonomics.
>
> But it does achieve something more than the workaround is able to: placing
> the m= application line at the top. The fact that the order of m= sections
> is more predictable reduces the risk of application bugs. I believe Google
> Meet had a serious regression a few years ago where the m= section order
> was something not what we expected causing the SDP parser to fail and the
> call to be disconnected. It's also a good idea that the m= application line
> gets tagged, avoiding risk of rejecting the whole bundle if the wrong m=
> section is rejected.
> On Thursday, January 22, 2026 at 3:20:18 PM UTC+1 Philipp Hancke wrote:
>
>> bear with me while I try to explain 15 years of history...
>>
>> Am Mi., 21. Jan. 2026 um 16:39 Uhr schrieb Rick Byers <
>> [email protected]>:
>>
>>> This is generally why we require explainers. Part of the point of this
>>> launch process is ensuring we've got the public material for web
>>> developers to understand the feature. Harald can you post an explainer
>>> somewhere which includes answers to Rob's questions?
>>>
>> Thanks,
>>>    Rick
>>>
>>> On Wed, Jan 21, 2026 at 9:30 AM Robert Flack <[email protected]>
>>> wrote:
>>>
>>>> The specification just says "alwaysNegotiateDataChannels defines a way
>>>> for the application to negotiate data channels in the SDP offer before
>>>> creating a datachannel".
>>>>
>>> I'm finding it a bit hard to reason about what this means for me as a
>>>> developer. I.e. when should I set it and how does this change the
>>>> connection?
>>>>
>>>
>> Lets say you create a RTCPeerConnection. Create an offer and call
>> setLocalDescription. Send it to your peer. You get an answer from the peer
>> and feed it into setRemoteDescription.
>> You might wonder why it does not result in a connection.
>>
>> Next, you create a datachannel. Your datachannel never opens because
>> after adding the datachannel (only the first!) you need to call
>> createOffer/setLocalDescription and call setRemoteDescription with the
>> answer.
>> "web developers" are notably confused:
>>
>> https://stackoverflow.com/questions/79737908/how-do-i-create-a-data-channel-in-webrtc-js-firefox-115-6-esr
>> Now your connection gets established. If you add another data channel you
>> do *not* need to do this negotiation again (thankfully; adding audio or
>> video tracks requires negotiation).
>>
>> In this case using the boolean flag is saving you one round of
>> negotiation.
>>
>> In many applications you do not know if you want to use a datachannel but
>> since you might, you preemptively create one and throw it away. You can see
>> that e.g. in Google Meet which calls
>>    pc.createDataChannel("ignored", {reliable: true})
>> It never uses that channel but it shows up in the getStats API every time.
>>
>>
>>> Does this mean that a data channel will be negotiated without calling
>>>> createDataChannel?
>>>>
>>>
>> The underlying SCTP association will be negotiated and can be used by
>> subsequent calls to createDataChannel without renegotiation.
>> This currently means exchanging four packets over the network (but that
>> upfront cost might go away later this year).
>> Everything is set up to just work.
>>
>>
>>> Will that requested datachannel result in a datachannel event or does my
>>>> code still need to call createDataChannel? If the latter, is this a
>>>> performance optimization?
>>>>
>>>
>> Part of it is a performance / ergonomics optimization. But behold, there
>> comes the 2011 backward compat angle below.
>>
>>
>>> Would I choose this when data is more important to connect first rather
>>>> than e.g. voice / video?
>>>>
>>>
>> Lets say you have a stream from getUserMedia with audio and video and
>> would also like to have a datachannel.
>> You call pc.createDataChannel and then iterate over the MediaStream's
>> getTracks and call pc.addTrack with each.
>> Because data channels were new and fancy in 2011 and backward compat was
>> a concern this results in SDP (due to rules in RFC 8829) which puts the
>> datachannel bits in the SDP after the audio and video bits (and in Safari
>> the order of audio and video bits depend on the lexical ordering of the
>> random uuid of the MediaStreamTrack) so the SDP looks roughly like this:
>>   v=0 (6 more lines)
>>   m=audio 9 UDP/TLS/RTP/SAVPF 63 111 (23 more lines) direction=sendonly
>> mid=0
>>   m=video 9 UDP/TLS/RTP/SAVPF 96 97 98 99 100 101 45 46 (63 more lines)
>> direction=sendonly mid=1
>>   m=application 9 UDP/DTLS/SCTP webrtc-datachannel (9 more lines)
>> direction=sendrecv mid=2
>>
>> Which is ok but you are probably scratching your head why this order does
>> not match the order of calls (or why in Safari you get
>> video-audio-application half the time)
>> https://www.rfc-editor.org/rfc/rfc8829.html#name-initial-offers
>> specifies this (search for "Lastly, ") but does not provide much of a
>> rationale. https://github.com/w3c/webrtc-pc/issues/735 has some 2016
>> rationale.
>> Now that seems like an odd choice but had little consequence until...
>>
>>
>>>> The linked tracking bug <https://issues.chromium.org/433898678>
>>>> appears to be a chrome-only bug report with a particular demo that doesn't
>>>> use the alwaysNegotiateDataChannels attribute and works on Firefox. I don't
>>>> quite understand what the new configuration extension has to do with the
>>>> original bug report.
>>>>
>>>
>> WebRTC added APIs that allow you to negotiate codecs at some point. This
>> can result in a situation where you end up with a incompatible set of video
>> codecs (see the bug) and if those happen to be first in the SDP this "m="
>> line will be marked as "rejected" and can not be used to convey the
>> transport information anymore.
>> At best this results in another round of negotiation to satisfy
>>
>> https://datatracker.ietf.org/doc/html/rfc8843#name-rejecting-a-media-descripti
>>
>> Now
>>   https://datatracker.ietf.org/doc/html/rfc8843#section-7.2.1
>> actually provides a recommendation on how to avoid this: you pick
>> something that is "unlikely to be rejected by the remote side".
>> The browser does not have an API that would lead to it rejecting a
>> datachannel m= line (which can happen for audio and video)
>> So it would be great to pick this datachannel m= line as "offerer tagged"
>> which avoids this headache.
>> But that requires (unless you want to risk more breakage) putting this m=
>> line first which is conflicts with the rules of RFC 8829.
>> The alternative is going through at least two rounds of negotiation
>> (which involves the remote end), the first datachannel only and then adding
>> media.
>> Hence you need a way to opt-in which is the boolean in the
>> RTCConfiguration.
>>
>> Would you like to hear more about why WebRTC is so weird? (a lot of this
>> boils down to "SDP is not an opaque blob)
>>
>>
>>>> On Wed, Jan 21, 2026 at 6:31 AM 'Philipp Hancke' via blink-dev <
>>>> [email protected]> wrote:
>>>>
>>>>> *tContact emails*
>>>>> [email protected], [email protected]
>>>>>
>>>>> *Specification*
>>>>>
>>>>> https://w3c.github.io/webrtc-extensions/#always-negotiating-datachannels
>>>>>
>>>>> *Summary*
>>>>> Implement
>>>>> https://w3c.github.io/webrtc-extensions/#always-negotiating-datachannels
>>>>>
>>>>> *Blink component*
>>>>> Blink>WebRTC
>>>>> <https://issues.chromium.org/issues?q=customfield1222907:%22Blink%3EWebRTC%22>
>>>>>
>>>>> *Web Feature ID*
>>>>> Missing feature
>>>>>
>>>>> *TAG review status*
>>>>> Not applicable (incremental addition to WebRTC)
>>>>>
>>>>> *Risks*
>>>>> *Interoperability and Compatibility*
>>>>> *No information provided*
>>>>>
>>>>> *Gecko*: No signal (
>>>>> https://github.com/mozilla/standards-positions/issues/1335)
>>>>>
>>>>> *WebKit*: No signal (
>>>>> https://github.com/WebKit/standards-positions/issues/599)
>>>>>
>>>>> *Web developers*: No signals
>>>>>
>>>>>
>>>>> *Will this feature be supported on all six Blink platforms (Windows,
>>>>> Mac, Linux, ChromeOS, Android, and Android WebView)?*
>>>>> Yes
>>>>>
>>>>> *Is this feature fully tested by web-platform-tests
>>>>> <https://chromium.googlesource.com/chromium/src/+/main/docs/testing/web_platform_tests.md>?*
>>>>> Yes, a WPT is part of
>>>>> https://chromium-review.googlesource.com/c/chromium/src/+/7080041
>>>>>
>>>>> *Rollout plan*
>>>>> Will ship enabled for all users
>>>>>
>>>>> *Requires code in //chrome?*
>>>>> False
>>>>>
>>>>> *Tracking bug*
>>>>> https://issues.chromium.org/issues/433898678
>>>>>
>>>>> *Estimated milestones*
>>>>> Shipping on desktop 146
>>>>>
>>>>> *Link to entry on the Chrome Platform Status*
>>>>> https://chromestatus.com/feature/5113419982307328?gate=6516338099093504
>>>>>
>>>>> This intent message was generated by Chrome Platform Status
>>>>> <https://chromestatus.com/>.
>>>>>
>>>>> --
>>>>> You received this message because you are subscribed to the Google
>>>>> Groups "blink-dev" group.
>>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>>> an email to [email protected].
>>>>> To view this discussion visit
>>>>> https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CADxkKiKuE79-xWUFiO9NCy%2Big9Pvn1oi6fVbkdomFLo4--bJQw%40mail.gmail.com
>>>>> <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CADxkKiKuE79-xWUFiO9NCy%2Big9Pvn1oi6fVbkdomFLo4--bJQw%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>>> .
>>>>>
>>>> --
>>>> You received this message because you are subscribed to the Google
>>>> Groups "blink-dev" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to [email protected].
>>>> To view this discussion visit
>>>> https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAJh39TPTxV5evdFUMnmfQE3aP2siNXNSZVpBr8wBitT56-nFCQ%40mail.gmail.com
>>>> <https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAJh39TPTxV5evdFUMnmfQE3aP2siNXNSZVpBr8wBitT56-nFCQ%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>>> .
>>>>
>>>

-- 
You received this message because you are subscribed to the Google Groups 
"blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAJh39TMsJtfPyCKB%2BUOiH_Hm%3DiMjMy2sM9uHk3FMSt7NRQgbzQ%40mail.gmail.com.

Reply via email to