record and play of probe events

[Rick Riolo]

Just a few comments on points glen raised:

1. I'm not sure where you are trying to go with
   the probe as "object violating".
   Probes are already in Swarm and (if I understand you) 
   are doing this object-violation" already.

   In fact, by *forcing* users to provide a 
   setMyVar method to make an IV probe-settable,
   I think that would get around some of the object-violation
   that the current implementation of probes do.

   But I don't want to go that far....its too inconvenient
   for all the users who don't want to do anything that fancy.

2. Re needing the same RNG seed to do a replay.
   That's going to be necessary anyway...I'm assuming
   all the initial state will be there for a replay.
   I'm just trying to get a way to catch exogenously caused
   changes of state that happen during the run so that
   those can be included in the replay.

   So I don't see how needing a RNG seed argues
   against time-based replays.

3. I think its important to allow users to 
   re-intervene during replays.  The replay "documention"
   just has to say "You'll get an exact replay if
   you just let it go, but if you change more things during
   the replay, all bets are off."  Including maybe having agents not
   be there that were expected to be there.
   I can't see that would be a surpise to any user.

   As I have my little semi-hack set up, you can intervene in
   the replays, and those get recorded right along with
   the original dynamic changes, so you can re-play the replay!
   (Of course its easy for me, since I don't have agents dying.)

In general, I think I can understand why glen wants to
get us to think about some of the deeper questions he is raising.
  
But I would like to argue for the 80/20 (or is that the 90/10?) rule...
You can get 80% of the functionality with 20% of the effort,
and doing that instead of waiting for that last 20% 
is worth a lot to a lot of people.
(Actually, I see the swarm deployment as having taken this approach.)

In this context, I'm arguing that by adding a simple hook
in Probes (one I thought was there, in fact!) will make
Probes much more open to customization by users, but
still gives the standard probe functionality with no need
for users to do custom programming if they don't want to.

Maybe the hook I want should be in some user interface
objects sitting between the user and probes...that's fine with me too.

- r

Rick Riolo                       address@hidden
Program for Study of Complex Systems (PSCS)
1061 Randall Lab     University of Michigan
Ann Arbor MI 48109-1120
http://pscs.physics.lsa.umich.edu/PEOPLE/rlr-home.html


                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ==================================