Clarify how to model streaming binary data

[handrews]

• Fixes Support for Streams in Responses #1576

Specifically, see #1576 (comment) for the TSC decision documented in this PR.

If this is accepted, I will add analogous information to 3.1.1 (after PR #3727 is resolved) and 3.2.0.

[mikekistler]

I have some concerns with this proposed change.

• It seems to address only streams of binary data, but streams (as I understand them) are also application for non-binary data (I think SSE is an example).

• The current change is being made in the "Data Types" section, which seems like the wrong place (to me) to describe how to represent streaming payloads. I think a better place for this information is in the Request Body and Response Body object, or maybe the Media Type object description.

[handrews]

@mikekistler This change is only intended to address the binary case because that's the part that the TSC agreed on in June of 2018 as documented in the linked issue. I'm just trying to resolve old issues and adding a previously-approved TSC decision seemed straightforward. You can see that I'm addressing JSON Streaming in PR #3735, and if you want other types of streaming addressed please feel free to open an issue / PR. But at least with binary and JSON streaming, the concerns are very different.

Adding the binary data streaming data to the data types section fits with adding binary handling to data types in general, which was already approved in PR #3187. If we're having a datatypes section about handling binary data, then I feel that the binary streaming should go with it.

[handrews]

@mikekistler while I'd rather not move the section that's already there, if you think this new addition should be separate from it entirely, that would be different. I'm also happy to consider moving the section as a separate issue/pr from whether to add this part. (I realize that might not have been clear from how I'd worded it in my previous (now edited) comment).

[iox666999]

闡明如何對流式二進位資料進行建模

[handrews]

@ralfhandl @mikekistler I added an explicit note about string data, made the whole thing a MAY, and did some other minor wording improvements.

I did look at the Response Object section, as I'm feeling more persuaded by @mikekistler 's argument about placement the more I think about it. But there's really nowhere obvious to put it there. I think we can revisit this if we want to make a more comprehensive statement on streaming responses.

[mikekistler]

Looks good! 👍