Re: [Adonthell-devel] Ideas from yesterday's meeting

[Alexandre Courbot]

> Hm, that is a problem. 
> 
> What if we limit the things a client does know to submaps? A submap is a
> rather small area (a room, or a limited outdoor area). Here cheating isn't
> really a problem, as you wouldn't gain much.

That's an idea. We might as well have bigger submaps - but that seems to
be a good compromise. On the other hand, each time the player changes a
submap, the client will have to request it to the server. It would be
more bandwith consuming than requesting the entire map, which can easily
be compressed when it's entire.

> Didn't my mail this morning come through? 

Sorry, missed it. I lost the habit of "high" traffic on the lists ;)

> You'd still do myobject->save() :). However, that does not write anything
> to disk yet. So what you reach is a seperation of normal program code and
> disk access code. As I said yesterday, good programs have three distinct
> vertical layers:
> 
>     GUI
>     -------
>     Program
>     -------
>     Disk/DB
> 
> As we want to achieve a clean design with v0.4, I think it's simply not
> feasible to have program logic and disk access stuff in the same class.

Uh, right. Absolutely right. I've read your other mail (have to read it
deeper to get all the details but I print it in order to read it
tomorrow), and the idea seems very good. A slight overhead as data moves
more, but on the other hand we could easily store the state of an object
and, as you said, transfert it to something else than a file (say, a
socket). But more to come on this matter tomorrow or tuesday, depending
of my available time! :)

Alex.

-- 
http://www.gnurou.org