Re: [MUD-Dev] Re: (no subject)

[JC Lawrence <claw#under,Eng.Sun.COM>]

On Sat, 27 Dec 1997 15:52:07 PST8PDT 
stad <Ola> wrote:

> JC Lawrence <claw#under,Eng.Sun.COM> wrote:

>> so far (say a week), with the excuse that quantumn randomity hides
>> the rest of the past from us through random "noise".

> Idea: What about providing some kind of selective world memory.
> That is, space/timeslices that has attracted a lot of attention will
> be available for more than a week.  (a replay of highlights feature)

The problem is that I can't do that technically with the current
implementation model (well I could, but it would be a bitch).  The
logging is implemented under the covers as context diffs, with the
result that I need a compleat chain of differences from the current
object state back to the past object state.  The diffs I store are
also not contextual to their objects. but to the basic types (atoms)
that objects are assembled from.  

The result is that to time watch a given object at a given instant in
the past, I more or less have to be able to recreate the entire game
universe as of that time in the past.  Big uglies.  This is why there
is a uniform death time for time watching -- you can't watch anything
older than one weel because the entire game world diff's are killed
for enything older than a week.

Hurm.

Actually have to re-examine this.  There may be a back route to the
same point to allow localaised preservation.

-- 
J C Lawrence                               Internet: claw#null,net
                                           Internet: coder#ibm,net
----------(*)                        Internet: jc.lawrence#sun,com
...Honourary Member of Clan McFud -- Teamer's Avenging Monolith...