Re: Map-related deja-vu

[Lincoln Peters <sampln at sbcglobal dot net>]

On Thu, 2004-09-30 at 17:26, Eric McDonald wrote:
> mskala@ansuz.sooke.bc.ca wrote:
> 
> > Is this a problem?  If so, we could easily put in a better seeding rule
> > and/or a better random number generator.
> 
> As I recall from looking at the 'util.c' code once upon a time, Xconq 
> uses a simple LCRNG. If we wanted something with a much longer cycle 
> length, we could go for a Mersenne Twister or some such. Of course, that 
> would not be suitable for cryptographic work, but neither is the current 
> one.

I don't know much about the logic (?) behind generating random numbers. 
But I did find it rather surprising when the map looked exactly like the
map I had played on just a few days ago.

I doubt that an algorithm suitable to cryptographic work would be
necessary for any game, but if I ever find myself playing the exact same
"random" game more than once after the random-number generation code is
upgraded, we'll know that the it still needs work.

> 
> But, better seeding would probably be a good first step.

Agreed.
I'm a bit surprised that nobody reported this before.  Perhaps that is
because:

1. Anyone else who experienced this problem didn't realize that the maps
were actually identical.

2. Anyone else who experienced this problem neglected to report it (this
doesn't seem as uncommon as we would like).

3. I just got lucky.


---
Lincoln Peters
<sampln@sbcglobal.net>

Between grand theft and a legal fee, there only stands a law degree.