Fixed #19, #20, #26: Require remove() for range removal

[jdsmith3000]

• Require remove() for all Range Removals
• Error on Duration Changes that need remove first
• Error on abort() during Range Removals

[jdsmith3000]

@wolenetz: Please review. Addresses issues #19, #20 & #26.

[paulbrucecotton]

@wolenetz; Can you please review this PR ASAP so we can get issues #19, #20 & #26 dealt with?

[wolenetz]

@paulbrucecotton @jdsmith3000 Yes - this is part of what I'm working through shortly.

[jdsmith3000]

@wolenetz: Might you review by EoW? Change implements our agreed upon solution. Addresses issues #19, #20 & #26.

[wolenetz]

Yes, I'm looking at this today - got tied up earlier this week in trying out upstreaming a new web-platform-test. Reviews and PRs on spec issues will be coming faster now from me.

[wolenetz]

@jdsmith3000 Looks pretty good. See my comments, above. Let me know when they're addressed and I'll do a final review and merge this if it l*tm. Thanks!

[wolenetz]

@jdsmith3000 Yesterday, I responded to your most recent comments in the line notes, above. I believe we're in agreement on the route forward. Please let me know if you have any further questions. I'd like to review again once you've addressed the line notes, above. Thanks!

[jdsmith3000]

@Wolenetz Rebased commit picks up these changes based on your feedback:

• Change Duration: used starting presentation timestamp in duration test
• Expanded note on using remove()
• Corrected indentation
• Updated dates

[jdsmith3000]

@Wolenetz Please review. I think I've picked up the changes we settled on after discussion. If so, let's finish this and close three issues.

[wolenetz]

Yep - now that I'm back from my brief vacation, I'm reviewing both this and https://github.com/w3c/media-source/pull/73 today.

[jdsmith3000]

@wolenetz I believe this is now ready to merge.

[wolenetz]

Looks good! Since the EME-discussion about the GH "green button" hasn't settled / I'm not sure I trust the button yet, I'll go ahead and merge this manually, shortly. Thanks for getting this updated, @jdsmith3000

Note to web authors and UA implementors: This spec change may break MSE players which previously depended upon implicit (and ambiguously defined) range removals on duration reductions which would truncate buffered media, and similarly might break players which attempt to abort() asynchronous range-removals. The workaround for the former is now a well-defined explicit remove(), and for the latter, there is no longer allowed any ability for an app to abort() range removals.

[plehegar]

I'm curious: does this change break existing UA implementations and, if yes, do we know the timeline for them to get aligned?

[jdsmith3000]

Yes, this will require implementations to update. The previous spec was not actually implementable however.

We need to discuss implementation updates in general, since this is not the only issue that could require updates.