When AMO/mkt runs in multiple data centers the current method of passing a JSON file of apps to the validator will not work. This needs to be fixed in two ways: 1) AMO/mkt needs to use the storage API for writing/reading the apps JSON file 2) the main validator function needs to accept a JSON object of apps, not a filename See this diff for clues on moving AMO/mkt to use the storage API https://github.com/mozilla/zamboni/commit/4220ef2c0dad92e693803c780c1b0a4e8a2218bd#commitcomment-1563995 These places are where AMO/mkt currently sends the JSON filename to the validator: https://github.com/mozilla/zamboni/blob/master/apps/devhub/tasks.py#L151 https://github.com/mozilla/zamboni/blob/master/mkt/developers/tasks.py#L162

Updated the validator to accept JSON strings and return JSON strings: https://github.com/mozilla/amo-validator/pull/171

Updated the pull request with updated docs, better tests, and the ability to pass the approved_applications as a dict.

Merged the pull: https://github.com/mozilla/amo-validator/commit/82d71f24c143f7b85957bb2296d9d3d4d05e9aaa Allen: You want to handle the updates to zamboni or do you want a hand?

whoops, too soon.

ok, here's what we can do now: in the AMO validator (https://github.com/mozilla/zamboni/blob/master/apps/devhub/tasks.py#L101) we can pass in dict of approved_apps. We still need to pass in a temporary filename for the XPI itself though. In the mkt validator (https://github.com/mozilla/zamboni/blob/master/mkt/developers/tasks.py#L107) we can call the new validate_apps() method and pass in a dict of the JSON manifest, no need for any temporary files.

Just as a correction: `validate_apps()` accepts a JSON string for its `data` parameter, not a dict. This way, we can return errors about the well-formedness of the JSON itself.