API design


     I'd like to discuss the issue that is closest to my heart, which is 
designing WebRTC for normal web developers, not telecom experts.

     I'll fire the opening salvo by recommending you watch this video: 

     As I mention to Cullen, this talk has shaped my professional 
career. Pay special attention to "Characteristics of a Good API", 
especially his explanation of the last bullet point :)

Thank you,

On 6/27/2013 10:19 AM, Gili wrote:
> Hi,
>     (If you'd like to respond to individual points, please start a 
> separate topic)
>     I'd like to start a discussion of issues that came up during the 
> WebRTC World conference (in sessions and while speaking with Dan 
> Burnett and Cullen Jennings):
>  1. Ending the VP8/H264 war: A proposal was made to mandate a
>     patent-unencumbered codec (whose patents have expired or are not
>     enforced) as mandatory and optionally upgrade to other codecs such
>     as VP8 or H264 depending on peer capabilities and personal
>     preferences. VP8 guys can use VP8. H264 guys can use H264. And if
>     the two camps need to chat with each other they can fall back on
>     H263. This gives you the flexibility of arbitrary codecs without
>     the need to do transcoding.
>  2. The WebRTC API needs to focus on normal web developers, not not
>     telecom experts: The conversation on this mailing list is unduly
>     skewed in favor of telecom experts which make up a tiny minority
>     of WebRTC end-users. We need to find a way to collect feedback
>     from the Javascript community at large in order to ensure that the
>     API facilitates their use-cases. The proliferation of WebRTC SDKs
>     for end-users (the conference was full of them) is a strong
>     indication that there is a gap to be filled.
>  3. Implementers vs End-users: The specification document has two
>     target audiences, implementers and end-users. We need to provide
>     implementers with a lot of low-level detail but make as little
>     guarantees as possible to end-users to leave the door open to
>     future change (without breaking backwards compatibility). We
>     discussed explicitly marking-up sections of the specification "for
>     implementers" or "for end-users" or separating the specification
>     into separate documents. We need to make it clear, for example,
>     that the specification does not make any guarantees regarding the
>     contents of the SDP token. Implementers need a detailed breakdown
>     in order to implement WebRTC 1.0 but end-users may not rely on
>     these details because the token might not even be SDP in future
>     versions.
>  4. SDP: Users should interact with the Constraints API instead of
>     SDP. It is true that there are some use-cases that are not yet
>     covered by this API (forcing you to manipulate the SDP directly)
>     but the plan is to address all these use-cases by 1.0 so users
>     never have to interact with SDP directly. "If your use-case is not
>     covered by the Constraints API, please tell us right away!"
>  5. Offer/Accept: There are plans to enable peers to query each
>     other's capabilities and change constraints (and as a result the
>     offer/answer) in mid-call.
>  6. Troubleshooting WebRTC: We need to do a better job diagnosing
>     WebRTC problems. We need a user-friendly application (run by
>     non-developers!) for quickly debugging network and microphone
>     problems (Skype does this), and allow users to drill down into
>     more detail if necessary. We also need programmatic access to this
>     API so WebRTC applications can detect problems at runtime and
>     decide (for example) to refund users who paid for a call that was
>     subsequently aborted due to network problems.
>  7. Use-cases, use-cases, use-cases: "Tell us what is wrong, not how
>     to fix it". You are a lot more likely to get traction for your
>     problems if you help us understand your use-cases then trying to
>     argue for change for its own sake. On the flip side for
>     specification editors, I encourage you to actively engage posters
>     (ask for these use-cases) instead of ignoring discussion threads ;)
>     I encourage other people who attended the conference to contribute 
> their own discussion points.
>     (If you'd like to respond to individual points, please start a 
> separate topic)
> Thank you,
> Gili

Received on Thursday, 27 June 2013 22:27:12 UTC