Comment 7 for bug 1031954

I think Mike was confused by a typo in my original note (I indicated a change was from 0.01ms to 3-4ms, while I meant seconds rather than ms... :)

I am not sure about the value that msgpack brings to the table in this case.

Apparently there are two parts with different characteristics to serialize:
     a. The device table and any other parameter we may save as part of teh ring apart from the arrays
     b. The very long arrays
So maybe we should treat the two differently.

As for a. any serialization will do - could be json or any other, or we could write our own. It is going to be fast anyhow. And we must use versioning as as Darrel indicated in case there is a need to extend the ring data one day.
As for b. It may be appropriate to write a very simple serialization code dedicated for the task. E.g. write each entry as an unsigned short (assuming the max number of allowed devices is limited). This can be written for example right after the the serialized representation of a.