Usage of pickles for persistence / serialization

[gutierrad]

Hi,

I would like to discuss about the usage of pickles for persistence / serialization.
I understand the advantages they provide, but I think it would be worth some time to verify that there are no inherent problems because of using them.
My main concerns come from:

• Cross-platform usage.
• Long term access to stored results.
• Compression of results.
  Possible alternatives would be:
• Human readable: JSON -> This are easy to create, though the files can be very big and sometimes the objects require specific serialization methods
• Binary: HDF / NetCDF -> For bigger objects, these may be the only option

Cheers,
D