Kedro is an open-sourced Python framework for creating maintainable and modular data science code.

Kedro

<https://bsky.app/profile/kylebarron.dev/post/3lsgtirm3i22x>

&gt; The primary existing Python specification used for object storage is <https://filesystem-spec.readthedocs.io/en/latest/|fsspec>, which defines a filesystem-like interface based around Python file-like objects.
&gt; However this presents an impedance mismatch: *object storage is not a filesystem* and does not have the same semantics as filesystems. This leads to surprising behavior, poor performance, and integration complexity.

I would love us to move away from our hard dependency on fsspec to be honest. been following Kyle's work for some time (see my earlier comment <https://kedro-org.slack.com/archives/C03QPDQTLBG/p1745253618125809>)

Agree on performance issue, question is what is the alternative?`fsspec` has always been a nice abstraction for the somewhat consistent interface, but sub-par performance. For example if I've use `botocore` directly few years ago, and I can easily get a few x setup for s3.

for now, I don't think there is a clear winner. AFIAK, pandas still use `fsspec`, while `polars` has its own thing in Rust and Spark has its own thing as well.

for me the performance is the least important thing. It’s about the API and afforddances 