Arvados

The code looks good. I appreciate that you moved the logic to the Ruby SDK.

So I intentionally corrupted the cached discovery document and got this error:

Warning: error reading cached discovery document /home/peter/.cache/arvados/discovery-86e60ccc4da91c726aa80dfb73fa9cbf.json: 795: 
unexpected token at ':"discovery#restDescription","discoveryVersion":"v1","id":"arvados:v1","name":"arvados","version":"v1","revision":"20131114",
"source_version":"1fae5d5","generatedAt":"2015-01-28T18:52:17+00:00","title":"Arvados API","description":"The API to interact with 
Arvados.","documentationLink":"http://doc.arvados.org/api/index.html","protocol":"rest","baseUrl":"https://qr1hi.arvadosapi.com/arvados/v1/",
... (and so forth for 1000s lines spewing the entire discover doc)

Consider not reporting an error in this case, because it is nonfatal and the user probably can't do anything with the information.

Does the Python discovery document cache code have the same robustness that the Ruby SDK will have with this branch?

Python appears to use a different caching scheme than Ruby, which means the discovery document gets stored twice?

Peter Amstutz wrote:

Consider not reporting an error in this case, because it is nonfatal and the user probably can't do anything with the information.

Yeah, with 100-line error messages that's definitely not going to fly. Updated in d59deccc.

Does the Python discovery document cache code have the same robustness that the Ruby SDK will have with this branch?

The Python SDK lets httplib2 handle caching. I haven't dug into its source but I think we should trust them to deal with it robustly. They have to handle users a lot more demanding than us.

Thanks.

LGTM