docs: document S3 and S3-compatible storage for the s3 filesystem

Author: goanpeca

docs/overview.md

@@ -19,6 +19,35 @@ Integrations with many systems and cloud vendors include (but not limited to):
 - Microsoft Azure Storage
 - Alibaba Cloud OSS etc.
 
+## S3 and S3-compatible object storage
+
+The `s3://` file system is registered when you `import tensorflow_io`. It reads
+the standard AWS credential and region environment variables
+(`AWS_ACCESS_KEY_ID`, `AWS_SECRET_ACCESS_KEY`, `AWS_REGION`), so it works with
+Amazon S3 out of the box:
+
+```python
+import tensorflow as tf
+import tensorflow_io as tfio  # registers the s3:// file system
+
+dataset = tf.data.TFRecordDataset("s3://my-bucket/train/shard-00000.tfrecord")
+```
+
+It also works with any S3-compatible object store, such as Backblaze B2,
+Cloudflare R2, MinIO, and others, by pointing the client at the provider
+endpoint with the `S3_ENDPOINT` environment variable. The client uses path-style
+addressing by default, so no extra configuration is required:
+
+```bash
+export AWS_ACCESS_KEY_ID=<access_key_id>
+export AWS_SECRET_ACCESS_KEY=<secret_access_key>
+export AWS_REGION=us-west-004
+export S3_ENDPOINT=https://s3.us-west-004.backblazeb2.com
+```
+
+The same `s3://` paths work for reading and writing checkpoints and SavedModel
+artifacts through `tf.io.gfile`, `tf.train.Checkpoint`, and `tf.saved_model`.
+
 ## Community
 
 * SIG IO [Google Group](https://groups.google.com/a/tensorflow.org/forum/#!forum/io) and mailing list: [io@tensorflow.org](io@tensorflow.org)