Re: [rtcweb] Sending of zero-length messages over data channels

["Makaraju, Maridi Raju (Raju)" <[e-mail]>]

Changing the send() API defeats the purpose of trying to provide full WebSockets compatibility.
<Raju> I understand the desire to make the API compatible to websocket API. But, unlike websocket TCP transport, the data channels the underlying transport (SCTP) has the flexibility for PPID, which can be utilized to achieve the same goal, with the added advantage of using the same approach for other use cases (like OOB data delivery).

IHMO, PPID_EMPTY approach is not straight forward either as at SCTP level you still have to send some dummy data (unless SCTP is enhanced) of some size (=1 or implementation dependent?), which is not obvious to the user. Will the ‘bufferedAmount’ be increased appropriately with the dummy data sent? Even if it is accounted for, the application has to know this and be mindful of the “mysterious” increase (1 byte or n bytes?) in ‘bufferedAmount’, something very unique to SCTP.

PPID_OOB approach is transparent to the user as user sends the data, so it is automatically accounted in ‘bufferedAmount’.

</Raju>

I think the PPID_EMPTY approach (probably PPID_EMPTY_TEXT and PPID_EMPTY_BINARY, to be precise) is a simple and complete solution.