During the last community sync we discussed possible interoperability
patterns between S3 and ofs/o3fs (Hadoop Compatible File System = HCFS)
Arpit suggested me to write down and formalize the exact proposal:
Current draft is here:
https://github.com/elek/ozone-notes/blob/master/content/design/s3_hcfs.md
(Will create a PR to make it easier to comment)
Existing discussion is under https://issues.apache.org/jira/browse/HDDS-4097
TLDR;
Normal (file-system path compatible) keys can be written from s3 and
read from ofs/o3fs without any problem/settings.
Tricky (not-compatible with file-system path) key names are not visible
from ofs/o3fs interface (this is the same as the S3A behavior), but can
be requested to normalize or reject instead (configuration).
Keys written from ofs/o3fs are always normalized and valid.
Please comment if you have any opinion,
Thanks
Marton
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
