Re: cheating & LOS

[philb@cats.ucsc.edu]

> 
> Cost under your scheme: 9 step * 11 MAPs/step = 99 MAPS
> 
> Cost under my scheme: The entire cross-corridor is mapped in (once) for  
> a cost of 3x11 MAPs.  For the remaining 6 steps only the 3 new squares  
> in the narrow corridor need to be mapped in for a cost of 6x3 MAPs.   
> Total cost: 51 MAPs.
> 
> So even in your worst case, the proposed protocol uses almost half the  
> number of MAPs which your proposed protocol requires !
> 

I think you're missing something here, Carl.
You're claiming LOS is BETTER than full transmission, in this case???
But you're forgetting the new right side of the corridor in yours, I think.

You cannot do better than full transmission of maps (with scrolling
extension) because under current rules, "blank" is just another face, or
"MAP".

Either method HAS to transmit a 11x11 grid, PLUS 11 "new" faces for each
step east after that. blank, or not blank, you still transmit them the
same way.

But the LOS method also has to blank approximately 5x9 image faces on the
right, and unblank 5x9 on the left.  Areas a + c (below) add to roughly
5x5, and so do b+d. 


 \_  a  ||  b  _/
    \_  ||  _/
      \ || /
--------  --------
   A         B
--------  --------
     _ /||\ _
  _ /   ||   \ _
 /    c ||  d   \


So for 9 steps, [ignoring protocol info]

LOS on server:  11x11 + 9x11 + 5x5 + 5x5 = 270 image faces
no LOS       :  11x11 + 9x11             = 220 image faces


   *******************************************************************


I think I picked up the rest of Jason's post, about using a "mask"

He proposes , instead of sending new faces, sending 11-word binary
viewing mask (?), that would change when neccessary. Each "word" would be
11 bits long , or whatever the width of the screen.

So for instance, if you were travelling down an unmarked corridor, the
mask would be

(more zeros here)
00000000000
00000000000
00000000000
11111111111
00000000000
00000000000
00000000000
(more zeros here)


The advantage to the mask idea being that, in "sparse" maps, where not
much is visible, you can compress 11 squares of darkness into a single
number.
I think this is getting too complicated, personally, and it's not _that_
much of a win, if at all.
(the trouble being that you HAVE to transmit 11 11-bit numbers when you
want to change the mask)


   ************************************************************

   
The "typeahead" movement thing, should be chucked, IMHO.
The client checking for key down is fine. But otherwise, it's too much
like making the player a robot. And that is dangerous to do in the server.
The client should be able to overide any time, but this might not happen,
netlag being what it is.

The client could implement the "typeahead" completely in itself with no
problems.