Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Lots of people here talking about S3, and yet:

> The system currently supports Azure Blob Storage and Google Cloud Storage.

I'd interpret that as either a hard "fuck you" to AWS, or a sign that the S3 API is somehow more difficult to use for this purpose.



The S3 API doesn't have the concurrency control primitives necessary to guarantee consistency in the face of concurrent writes.

I wrote a distributed lock on Google Cloud Storage. https://www.joyfulbikeshedding.com/blog/2021-05-19-robust-di... During my research it was quickly evident that GCS has more concurrency control primitives than S3. Heck S3 didn't even guarantee strong read-after-write until recently.



While I'm sure you're right, this VFS implementation doesn't use any of that stuff. They just handwave concurrent writes away by saying the application has to ensure a single writer on its own. An S3 implementation appears to be planned; it's mentioned in the comments in the bcvmodule.c file.


Ah, I'm guessing that's the same reason why many "cloud storage as a backend" stuff (e.g. Terraform/Vault storage backends, Delta.io backends) require DynamodDB + S3


I thought it was a sign of who paid for the feature. I have no insight as to whether or not this is true, but leaving out S3 made me assume either Google or Microsoft (or a third party heavily tied to one of these clouds) paid for this to be added.


I highly suspect it would be the latter because if you just want to "fuck" S3 then you could just list Minio, Swift and Ceph RADOSGW support instead.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: