Why have separate validate_op and apply?

[Kiiyya]

Wouldn't it make more sense for apply() to return a Result<(), Validation> instead, and not have any validate_op at all?

[davidrusu]

There are instances in real world applications where we want to validate an Op without applying it. For example, in the byzantine-reliable-broadcast protocol we separate out the validation phase from the applying phase: https://github.com/maidsafe/brb_dt_orswot/blob/master/src/brb_orswot.rs#L99-L131

2-phase commit follows a similar pattern of PREPARE followed by COMMIT. It's a common pattern across distributed algorithms so we want to be flexible in our API's here to support these cases rather than monolithic functions that do it all.

[Kiiyya]

Ah, looks like I have to do some reading on byzantine reliable broadcast then. Thanks for the answer!
I'll accept your answer once GitHub fixes its bug...