[Home] [Groups] - Message: [Prev in Group] [Next in Group]

nu.kanga.list.mud-dev

12452: Re: [MUD-Dev] distributed objects

[Full Header] [Plain Text]
From: "David Bennett" <ddt@discworld.imaginary.com>
Newsgroups: nu.kanga.list.mud-dev
Date: Tue, 15 Feb 2000 13:10:59 -0800
References: [1]
Organization: Kanga.Nu
On 2/15/00, at 10:50 AM, Vijay Weasel Prabhakar wrote: 

>	Not really, at least not in a combat system.  You can't tell what
>a random number is going to be unless you provide the seed, so it's tough
>to say that two rounds of combat are going to turn out the same way on two
>different clients.  

It is easy enough to create your own random number generators to do this
though.

>	I think the eventual objective is to reduce the requirements on
>the server's side by pushing work off to the clients.  The redundancy will
>really go against this objective.  You might as well be doing everything
>on the server side.  I think you could always figure out that your object
>is yourself.  It wouldn't take too much looking to be able to identify a
>few traits that establish your identity and then search for those traits.
>Random spot checks would not really work unless you were keeping up with
>the client constantly in which case you're doing as much work as you were
>before.  

True, but you need to reduce the effort in a way that is secure and works.
The idea suggested here, I belive, means that you don't trust clients much
when they first connect and double check everything they do.  Then you
slowly back off on the checks, doing less and less double checking.  This
would still reduce the load on the server while making the system more
secure.  (You just throw off people who get marked as untrustworthy, or
just stop their character effecting anything).

Good luck!
David.



_______________________________________________
MUD-Dev maillist  -  MUD-Dev@kanga.nu
http://www.kanga.nu/lists/listinfo/mud-dev