Replaying internal free parameters (from IRC)

[Hal Murray]

> Yes.  But that will break if any of
> the internal free parameters
> change.  For a simplre example:
> default polling interval.

That doesn't break things.  It just complicates them.

You will have to collect new data that covers the new defaults.  There is no 
way around that.

If you want to replay the old data files, I think it will take two changes.

First, the code will have to change slightly.  The current code has things 
like:
  #define NTP_MINDPOLL      6
  my_node->ctl.minpoll = NTP_MINDPOLL

The new code will have to have a variable that replay can change.  It gets 
initialized to the new NTP_MINDPOLL.  Replay will have to set it back to the 
old default before anybody looks at it.

Then, record mode will have to put a initial-conditions line at the start of 
each log file to record the values that change.  Since we can't/won't 
anticipate all possible changes, the replay side will have to fill in the old 
values that didn't get written out from built in versions of the old 
constants.  (It may be simpler to have a version number for the collected 
parameters.)

That sort of kludgery is probably worthwhile if we have a collection with 
some obscure cases in it.  For example a leap second happening or an obscure 
buggy response from a remote server.



-- 
These are my opinions.  I hate spam.