severity 311307 wishlist
thanks

On Mon, May 30, 2005 at 05:34:27PM -0500, Adam Majer wrote:
> This is supposedly an "action" game and yet it uses TCP connections.
> Games should use UDP connections *always*. TCP is unreliable (in games) and
> undesirable protocol to use. TCP is for streaming data, like files,
> audio, movies, etc..
> 
> The only disadvantage of UDP over TCP in games is that it is more
> difficult to write an engine around UDP since packet order must be
> maintained in the engine. I don't think that is an excuse!
> 
> - Adam
> 
> PS. Important because it makes the entire game experience worse. For
> other examples of games where TCP "wrecked it" see xshipwars.

Hello Adam.

Did you play netpanzer at all? I played it many times even with 20 players
on map and there were NOT any problems with speed.

Here are some quotes from my talk with upstream maintainer:

12:18 <    MatzeB> TCP is unreliable
12:18 <    MatzeB> that guy has no idea what he is talking about
12:18 <    MatzeB> the whole idea about TCP is that it is reliable
12:19 <    MatzeB> (you use UDP if you can accept unreliablilty...)
12:19 <    MatzeB> before he has hard facts or knows what he is talking about I 
                   won't discuss that
12:19 <     fEnIo> it is possible to use udp and have reliability but you have 
                   to implement it yourself... so in fact you end up with tcp 
12:20 <    MatzeB> well you don't end up exactly with tcp depending on your 
                   needs
12:20 <    MatzeB> as you could decide to not resend certain packets when 
                   they're missing
12:20 <    MatzeB> (like updating panzer positions, when you send 5 updates and 
                   miss an update in between in won't matter)
12:20 <     fEnIo> anyway... I'll lower severity of this bug and try to discuss 
                   with him
12:20 <    MatzeB> but netpanzer runs well enough with TCP IMO
12:21 <    MatzeB> I don't have any figures or experiences that TCP has any bad 
                   effect here
12:21 <    MatzeB> so I won't do all the trouble implementing an UDP protocol
12:21 <    MatzeB> I already did that in other games (planeshift) and know what 
                   a pain it is
12:21 <    MatzeB> and how many bugs you can have in it
12:22 <    MatzeB> just copy&paste this irc log if you want
12:22 <    MatzeB> and close that bug
12:22 <     fEnIo> well it could probably use less bandwidth with udp but I 
                   also think netpanzer works great with tcp
12:22 <    MatzeB> it also won't take less bandwith with udp
12:22 <    MatzeB> (you might save a few bytes, but nothing to talk about)
12:22 <     fEnIo> it hardly depends on what do we really needs
12:22 <    MatzeB> all you could win is improved latency over bad connections
12:23 <     fEnIo> ok nevermind... I'll paste this talk to bugreport
12:23 <     fEnIo> all in all if he wants to move to udp then patches are 
                   welcome ;)
12:23 <    MatzeB> probably :)
12:24 <     fEnIo> ok thanks for comments
12:24 <    MatzeB> if he can show me that it really works better with UDP and 
                   has a clean and bugfree implementation then yes
12:24 <    MatzeB> np
12:26 <    MatzeB> apart from that I don't understand why people are filling 
                   such a thing as debian bugreport...
12:26 <    MatzeB> I doubt any packager would patch the protocol to use UDP 
                   instead of TCP just for debian...
12:40 < ziberpunk> MatzeB, he's not saying tcp is unreliable 'per se', but for 
                   games
12:41 <    MatzeB> and still that doesn't make any sense
12:41 <    MatzeB> it's still reliable
12:41 < ziberpunk> MatzeB, well, it's true that most games uses UDP
12:41 < ziberpunk> MatzeB, yes, that's true
12:41 <    MatzeB> I know
12:41 < ziberpunk> MatzeB, I agree that netpanzer works well with TCP
12:41 <    MatzeB> but this report doesn't contain any facts why TCP is bad for 
                   netpanzer
12:41 <    MatzeB> it's just unbacked statements
12:41 < ziberpunk> MatzeB, yes, that's the point
12:42 < ziberpunk> MatzeB, performance data or bottlenecks would give more 
                   credibility :)
12:43 <    MatzeB> or just a few empirical statements ala the game lags like 
                   crazy for me
12:43 <     fEnIo> I'll just decrease severity to wishlist
12:43 <    MatzeB> (or even better with measured ping to the server and some 
                   estimations or comparisons with other games and similar
ping)
12:45 < ziberpunk> MatzeB, yes, the statements seems based on experiences with 
                   other games, and that doesn't worth rewriting the net code


So as you can see there won't be probably rewrite of net code.
Submitting it with severity "important" doesn't seem for me as reasonable
either so I've just decreased it to wishlist.

regards
fEnIo

-- 
  ,''`.  Bartosz Fenski | mailto:[EMAIL PROTECTED] | pgp:0x13fefc40 | irc:fEnIo
 : :' :       32-050 Skawina - Glowackiego 3/15 - w. malopolskie - Poland
 `. `'           phone:+48602383548 | proud Debian maintainer and user
   `-          http://skawina.eu.org | jid:[EMAIL PROTECTED] | rlu:172001

Attachment: signature.asc
Description: Digital signature

Reply via email to