This is very disappointing. I'm kind of amazed that they think this is a good business decision when there are massive numbers of paying customers with mountains of python code using boto3. I imagine much of that code uses the resource interfaces since they're easier to use.
This isn't some fringe feature they're deprecating; the official documentation uses these interfaces in its examples; they're literally telling you that's the correct way to use the library and then the next moment they're telling you it's not.
The AWS Python SDK team is no longer planning to support the resources interface in boto3. Requests for new changes involving resource models will no longer be considered, and the resources interface won't be supported in the next major version of the AWS SDK for Python. The AWS SDK teams are striving to achieve more consistent functionality among SDKs, and implementing customized abstractions in individual SDKs is not a sustainable solution going forward. Future feature requests will need to be considered at the cross-SDK level.
Unlikely, CDK is for infra provisioning, while people can use boto for that a heap of logic above the deployment layer is depending on this feature at the app . Multipart upload is handled via this, iterating over S3 keys or DynamoDB items is probably the most famous example that is inside of this abstraction.
Every service has their own snowflake way of handling iterating over large amounts of keys / resources. I can see it as a massive pain for them to upkeep that within boto. Ideally they would have done the opposite and introduced these abstractions to the other SDKs, but that would be a little too customer obsessed.
Being slightly dramatic but if the plan is to use Smithy make all the SDK as dumb as the lowest common denominator across the SDKs, I might as well use curl and interact with the API directly now that curl supports Sigv4 https://curl.se/libcurl/c/CURLOPT_AWS_SIGV4.html
By "the next major version of the AWS SDK for Python" they must mean boto3 2.0.0, but I don't see any timeline or roadmap – the current version is 1.26.53 and it's been 1.x.x since July 2015. When will this actually be removed?
This isn't some fringe feature they're deprecating; the official documentation uses these interfaces in its examples; they're literally telling you that's the correct way to use the library and then the next moment they're telling you it's not.