Is it possible to fail *and* retry? Or do we need to really give up once we report an error back? Or is it ok to just delay error reporting indefinitely?  Stalling and prebuffer statistics is probably better indicators of how good things are going anyways.

I don't follow your suggestion. Can you elaborate?

It's not possible to issue a network read error and continue the pipeline today. And, I think doing that would be hard too since we'd be relying on ffmpeg to handle the failure in a recoverable way.

I think isolating the retry logic within a place where we can ensure it's well tested is likely to be more successful.

I'm more worried about what we report to the javascript. I was wondering if we could tell the javascript that playback failed, but then change our minds later if the network becomes available again. The pipeline would continue to retry the connection, possibly forever.

Hmm, no I think once you report an error that has impacts on the state machine exposed to JavaScript -- it has no concept of a recoverable error at present AFAIK.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/018c484fd85b3a64ef1037191ca513b389fa21ff

commit 018c484fd85b3a64ef1037191ca513b389fa21ff
Author: hubbe <hubbe@chromium.org>
Date: Fri Mar 11 21:02:59 2016

Improve retry support for media network loading.

Retry 30 times in 30 seconds instead of 3 times in 750ms.
The retries are spaced so that the first retry is 250ms, second is 300ms, then 350ms, etc.

BUG= 592703 

Review URL: https://codereview.chromium.org/1777153002

Cr-Commit-Position: refs/heads/master@{#380723}

[modify] https://crrev.com/018c484fd85b3a64ef1037191ca513b389fa21ff/media/blink/buffered_data_source.cc
[modify] https://crrev.com/018c484fd85b3a64ef1037191ca513b389fa21ff/media/blink/buffered_data_source.h
[modify] https://crrev.com/018c484fd85b3a64ef1037191ca513b389fa21ff/media/blink/buffered_data_source_unittest.cc
[modify] https://crrev.com/018c484fd85b3a64ef1037191ca513b389fa21ff/media/blink/multibuffer_data_source_unittest.cc
[modify] https://crrev.com/018c484fd85b3a64ef1037191ca513b389fa21ff/media/blink/resource_multibuffer_data_provider.cc
[modify] https://crrev.com/018c484fd85b3a64ef1037191ca513b389fa21ff/media/blink/resource_multibuffer_data_provider.h

Your change meets the bar and is auto-approved for M50 (branch: 2661)

Please merge your change to M50 branch 2661 if you think it is a safe merge. If merge happens by Monday (03/14) 5:00 PM PST, we can take it in for next week beta. Thank you.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/dc4b45b89b72a24bd668cdfd3f5e33aec091c1df

commit dc4b45b89b72a24bd668cdfd3f5e33aec091c1df
Author: Fredrik Hubinette <hubbe@google.com>
Date: Mon Mar 14 22:43:37 2016

Improve retry support for media network loading.

Retry 30 times in 30 seconds instead of 3 times in 750ms.
The retries are spaced so that the first retry is 250ms, second is 300ms, then 350ms, etc.

BUG= 592703 

Review URL: https://codereview.chromium.org/1777153002

Cr-Commit-Position: refs/heads/master@{#380723}
(cherry picked from commit 018c484fd85b3a64ef1037191ca513b389fa21ff)

Review URL: https://codereview.chromium.org/1796973004 .

Cr-Commit-Position: refs/branch-heads/2661@{#227}
Cr-Branched-From: ef6f6ae5e4c96622286b563658d5cd62a6cf1197-refs/heads/master@{#378081}

[modify] https://crrev.com/dc4b45b89b72a24bd668cdfd3f5e33aec091c1df/media/blink/buffered_data_source.cc
[modify] https://crrev.com/dc4b45b89b72a24bd668cdfd3f5e33aec091c1df/media/blink/buffered_data_source.h
[modify] https://crrev.com/dc4b45b89b72a24bd668cdfd3f5e33aec091c1df/media/blink/buffered_data_source_unittest.cc
[modify] https://crrev.com/dc4b45b89b72a24bd668cdfd3f5e33aec091c1df/media/blink/multibuffer_data_source_unittest.cc
[modify] https://crrev.com/dc4b45b89b72a24bd668cdfd3f5e33aec091c1df/media/blink/resource_multibuffer_data_provider.cc
[modify] https://crrev.com/dc4b45b89b72a24bd668cdfd3f5e33aec091c1df/media/blink/resource_multibuffer_data_provider.h