datalad: choice of archive method and parameters for add-archive-content

[bpinsard]

Hi, just found this repo and I have a question regarding a datalad usecase.
I need to store dicoms archives (1 per MRI session) in cloud buckets before being indexed with datalad add-archive-content and I have control on the choice of archiving software.

I am looking for an archive format and parameters that would:

• compress a minimum the redundant info (eg dicom headers).
• allow (pseudo)-random access: not necessarily 0-fee access, but better than decompressing the full archive to get a single file.
• is well and long-term supported, available/easy-installable on most distribution.

I cannot find any compelling information to help me make that choice and I guess this has already been explored.
Thanks for your insight.