Remove "SendStream" Web IDL interface #307
Conversation
Have it a spec-only concept for streams created by the "create a SendStream" procedure. This is for #306.
|
@ricea, can you take a look? I'm leaning to pretending the internal slots to be attached to |
|
I'm concerned that implementers may feel they actually have to create a way to attach extra data to a WritableStream. But maybe we could make it clearer in the note that this is just an abstraction. Maybe we could call it a "send stream" instead of a SendStream to reduce the expectation that it would literally appear in the implementation. |
|
Added a note. |
index.bs
Outdated
| @@ -1117,6 +1114,11 @@ A {{SendStream}} has the following internal slots. | |||
| <tbody> | |||
| </table> | |||
|
|
|||
| Note: These internal slots need to be attached to {{SendStream}} (which is a kind of | |||
| {{WritableStream}}s) conceptually, not phisically. For example, implementers MAY have a weak hash | |||
There was a problem hiding this comment.
Maybe the "s" in {{WritableStream}}s isn't needed.
Spelling: "physically".
I don't think you should use MAY within a note section.
In general, I would prefer to recommend an implementation strategy where the slots are owned by an object which is owned by the algorithms.
Something like
For example, an implementation could place them on an object which is
referenced from writeAlgorithm, closeAlgorithm and abortAlgorithm.
There was a problem hiding this comment.
Fixed nits: thank you!
In general, I would prefer to recommend an implementation strategy where the slots are owned by an object which is owned by the algorithms.
I can do that: Just to confirm, are you fine with the recommendation with https://w3c.github.io/webtransport/#send-stream-STOP_SENDING?
There was a problem hiding this comment.
I'm confused. That algorithm references a "remoteCanceled" promise which isn't defined anywhere.
There was a problem hiding this comment.
Ah sorry I forgot to remove the step. I'll do that separately.
What I'm talking about here is the reference to the [[Transport]] internal slot at the first step.
- Let transport be stream’s [[Transport]].
There was a problem hiding this comment.
I think that's okay. It's not ambiguous.
Have it a spec-only concept for streams created by the
"create a SendStream" procedure.
This is for #306.
Preview | Diff