[Freeciv-Dev] Re: (PR#3648) Superfluous tiles bug

["Raimar Falke" <rf13@xxxxxxxxxxxxxxxxx>]

On Fri, Mar 07, 2003 at 05:03:10AM -0800, Arnstein Lindgard wrote:
> 
> I think this patch here will significantly reduce lag when a player
> reconnects. It should probably be applied to the pubservers.
> 
> I set up a client to study incoming packets in the hope of optimizing
> the network code itself, failed miserably, but found some redundancy
> that seems not to increase the systems robustness.
> 
> When a dead player reconnects, the number of PACKET_TILE_INFOs is
> reduced by 50% because each tile was blessed with to identical packets.
> When a living player reconnects, the gain seems to be ~43%.
> 
> The other patch deals with Apollo. Seems the whole damn map is sent
> every turn, and in duplicate because of the first bug! In function
> do_apollo_program() it seems superfluous to send tile and city info,
> since it's taken care of elsewhere in the code after map_know_all()
> is called. You still get updates when a fogged city changes.
> 
> I tested by having the client check if a packet actually contained
> new information. In the case of Apollo, 4000 out of 4006 packets each
> turn were wasted (default mapsize 80 x 50 = 4000).

This is a known problem. See
http://lists.complete.org/freeciv-dev@xxxxxxxxxxx/2001/10/msg01015.html.gz
for an early post. The most current solution to tackle this problem is
the delta patch from Nov 2002 (sorry no URL since the archive doesn't
contain the date).

        Raimar

-- 
 email: rf13@xxxxxxxxxxxxxxxxx
 "A common mistake that people make, when trying to design 
  something completely foolproof is to underestimate the 
  ingenuity of complete fools."
    -- Douglas Adams in Mostly Harmless