23743 2013-11-06 17:47:39 +0000 Split the functionality to read exact N bytes into readExact() method 2013-12-05 19:32:19 +0000 1 1 1 Unclassified WebAppsWG Streams API unspecified All All RESOLVED FIXED P2 normal --- 1 tyoshino tyoshino mike public-webapps public-webapps-bugzilla oldest_to_newest 95875 0 tyoshino 2013-11-06 17:47:39 +0000 The purpose of size argument of read() are - allow user to get notified only when the specified size of data is available and get it as one non-fragmented ArrayBufferView - flow control. I.e. not to tell the producer to generate, receive from wire, etc. too much Current interface doesn't allow flow controlled consumption without specifying size N which means until N bytes become available, we don't get the result. read() without argument helps for this case, but it's unclear how many bytes read() pulls. If we differentiate exact read and non-exact read mode, we can give read() clearer semantics - read() in exact mode: Prohibit? - read(N) in exact mode: I can accept up to N bytes. tell me when exactly N bytes are available - read() in non-exact mode: tell me when any amount of data is available - read(N) in non-exact mode: I can accept up to N bytes. tell me when any bytes are available 95969 1 tyoshino 2013-11-07 18:23:40 +0000 Instead of bothering readType, have a separate method named readExact. 95974 2 tyoshino 2013-11-07 21:14:27 +0000 Done in preview ver https://dvcs.w3.org/hg/streams-api/raw-file/tip/preview.html 96276 3 tyoshino 2013-11-14 01:47:08 +0000 Looks like we lack say pipeExact() which is necessary for propagating pull requests. 97207 4 tyoshino 2013-12-05 19:32:19 +0000 read(N) in non-exact mode has been realized by pullAmount. pipeExact() is unnecessary. Available space of the destination is propagated by pipe() by setting pipePullAmount to the value.