[Freeciv-Dev] Re: non-existent technologies

[Raimar Falke <rf13@xxxxxxxxxxxxxxxxx>]

On Fri, Jan 31, 2003 at 10:18:42PM -0600, Mike Kaufman wrote:
> On Fri, Jan 31, 2003 at 08:22:17PM -0500, Eric S. Raymond wrote:
> > Jason Dorje Short <vze49r5w@xxxxxxxxxxx>:
> > > There are all sorts of other unnecessary ruleset dependencies in the 
> > > savegames, for instance units are stored by ID rather than by name. 
> > > This can be improved on greatly by avoiding using an "index" when a name 
> > > can be used instead.  But it will still break if a unit is removed, for 
> > > instance.  To get full compatability either a capstring is needed to 
> > > match ruleset and savegame, or the savegame should store all needed 
> > > inforamation about the ruleset it was originally played with.
> > 
> > I am in favor of any move away from magic numbers to readable tokens.
> > The marginal extra space cost of storage and the marginal time cost of name
> 

> this is crap. marginal? I think not. you're talking about adding
> something like 150k to each savegame (to store ruleset info). This
> is ludicrous. We already store the name of the ruleset that was used
> to save the game. That should be good enough. What happens if I
> don't happen to have the same ruleset that the savegame was saved
> with? Should it simply refuse to load, or should it try the best it
> can? Even with a system that stores names rather than index, you're
> still screwed if you change the rulesets too much anyway. If you
> don't include all the ruleset data, then this is almost a worthless
> change, and including all the ruleset data is far too expensive.

I agree. A safe solution either needs to save all the ruleset data or
a hash/fingerprint of all (used) ruleset data. The latter may be
possible looking at the frequency of changes of the ruleset.

        Raimar

-- 
 email: rf13@xxxxxxxxxxxxxxxxx
 "Despite all the medical advances of the 20th century, the mortality 
  rate remains unchanged at 1 death per person."