1256451 - make it easier to recover from a bad service worker

Status: NEW

(Core :: DOM: Service Workers, defect, P3)

Description

[Ben Kelly [:bkelly, not reviewing]]

Recently I encountered a site in the wild that had a bad service worker.  I was getting "content corrupted" consistently.  They did push a fixed service worker script quickly, but it took me a while to see the fix.  I think because of the http cache holding the old bad script.

We should consider some measures to improve this situation.  Some ideas:

1) When we see a failed interception leading to "content corrupted", bypass the http cache when firing the navigation update.
2) Track number of failures and after some threshold unregister the service worker.
3) Like (2), but instead just bypass the service worker until an update occurs.

I think (1) is probably something we should do no matter what.  Not sure about (2) or (3).

Note, I believe chrome is implementing something like (3).