Api2.1

[abudaan]

I have drafted a new API and I am open to suggestions from the community.

You can find the draft in the change.log

Thanks!

[headlessme]

I have just read through the migration document. I had a query on some of the design decisions for the new API, particularly around how the bucket name is used. It looks as though I can specify the bucket name in the storage URL still (which is great, that's how we want to use it as it keeps our config very simple as just a single url), but then the strange bit, it looks like I also have to re-specifcy it in every call on the storage adapter? This seems a bit weird and unfriendly as I already specified it. Maybe I've misunderstood the change?

I'd expect that if I've specified the bucket in the URL of the storage setup then that's the ONLY bucket I can access through that particular storage instance (credentials are likely tied to that bucket anyway). At the very least it should be used as the default if I omit the bucket name in later storage API calls. If I don't specify the bucket in the storage constructor then it makes sense to require it on every later API call.

Secondly the changes in the URL format, was there a reason behind diverging away from the more known s3://buckname/path/to/object type URIs that the AWS cli supports?

[headlessme]

Both GCP and AWS seems to use a similar URI format for bucket identification:

GCP: gs://bucketname/path/to/object
https://cloud.google.com/storage/docs/gsutil

AWS: s3://buckname/path/to/object
https://docs.aws.amazon.com/cli/latest/userguide/cli-services-s3-commands.html#using-s3-commands-managing-objects-param

I understand extra params are required such as region, credentials etc... but the v1 formats handled that nicely.