Re: [Webtransport] A comment on draft-kinnear-httpbis-http2-transport-02

And to the original question about JS API work, I’m not aware of any, however there is a (not-yet-publicly-available) implementation of a native interface that does support this as a general abstraction. I think one of the key goals in WebTransport is going to be making sure that what we’re defining here has a clear boundary between: 

(a) Protocol mechanisms that are expressed on the wire
(b) Mappings to APIs that provide better abstraction, both Web and native.

I can see several possibilities for (b), but we should likely discuss that much more.

Thanks,
Eric

> On Nov 19, 2019, at 6:31 PM, Eric Kinnear <ekinnear@apple.com> wrote:
> 
> Agreed, I think these are somewhat separate concepts. I think draft-kinnear-httpbis-http2-transport is working towards extending HTTP/2 for bidirectional transport *beyond* WebSockets, if you’re intending to send WebSocket data on a stream, you’d use “websocket”, not “bytestream”, and you really probably should register another :protocol value to use for your protocol. The latest rev of the draft here indicates that if you’re going to use “bytestream” you must have negotiated what’s on the bytestream in a different manner. 
> 
> Lots more to unpack here, and always happy to chat, but that’s the short version of “bytestream” =/= “websocket”.
> 
> Thanks,
> Eric
> 
> 
>> On Nov 19, 2019, at 1:56 PM, Victor Vasiliev <vasilvv=40google.com@dmarc.ietf.org <mailto:vasilvv=40google.com@dmarc.ietf.org>> wrote:
>> 
>> On Tue, Nov 19, 2019 at 12:54 PM Martin J.. Dürst <duerst@it.aoyama..ac.jp <mailto:duerst@it.aoyama.ac.jp>> wrote:
>> The distinction between binary and text datagrams, and the limitation of 
>> text datagrams to UTF-8, considerably simplifies the creation of 
>> datagram-based protocols and increases interoperability.
>> 
>> I wonder why this wasn't included in the Webtransport draft, and hope 
>> that it can be added.
>> 
>> Hello Martin,
>> 
>> WebTransport does not make any assumptions about the nature of the payload it transmits, that is to say, all of the payload is assumed to be binary.  That is consistent with the general way the similar recent API has handled this (e.g. streaming fetch), as well as the design of underlying protocols.  TextEncoderStream/TextDecoderStream can be used to easily transform one into the other.
>> 
>> I'm especially confused about this being related to the datagrams; almost all datagram protocols I've seen are binary-based, since ~1200 bytes is typically too small for text-based protocols.
>> 
>>   -- Victor.
>> -- 
>> Webtransport mailing list
>> Webtransport@ietf.org <mailto:Webtransport@ietf.org>
>> https://www.ietf.org/mailman/listinfo/webtransport
> 