...but you can't tuna fish
There have been a lot of people complaining about latency tuning problems, dealing with people on 14.4K modems, and speed issues in general. In fact, some servers outright ban 14.4K modems yet will eagerly accept 28.8K speeds.

Juri Munkki refutes such misconceptions and clears up the whole issue in a posting to a USENET newsgroup:

From: jmunkki@alpha.hut.fi (Juri Munkki)
Newsgroups: comp.sys.mac.games.action
Subject: Avara Game Latency Tuning

In article <edremy-2409961150470001@nnrp.usc.edu>
edremy@chem1.usc.edu (Eric Remy) writes:
>One possible suggestion: could you add a "Max transmission delay" box to
>the server options dialog?

It's not that easy...well the biggest problem is that you might reject
some of those T1 players. As you said, the initial "guess" is often
wrong. It's based on the first few packets and sometimes they travel a
lot slower than once the route has been "established".

>Reason: Even though you put "T1 or better only please" [1] in the comments
>section, modem folks still join these games.  It's a pain to get six
>people together, start a game, then find that the game crawls.  You then
>pause, look for the guy with the bad net stats, and ask them to leave. 
>It's awkward as well: I hate to spoil someone's fun.  It would be far
>easier to just deny access to anyone more than 200ms roundtrip away.

I do have something to offer you...there's a secret "ping" within Avara.
You click on the net statistics (roundtrips & latencies) and it will
send a ping packet to bounce back and forth between that machine and you.
Click on it a few times quickly and the roundtrip estimates will usually
converge quickly to what they should really be. (It doesn't work, if
the other player is active in a game...this is so that you can't disturb
players with pings.)

It's then pretty easy to ask players with high latency to leave the game
or kick them out. Please try to be polite though.

>It seems that the net stats posted before a game starts are occasionally
>wrong: I've had people listed at LT 13 who were on T1s and folks at LT2
>who were on 14.4s- the stats correct themselves quickly once the game
>starts though.

14.4 bashing again?

I did some testing a few days ago and it confirms my previous experience.
I used a unix host close to my PPP server to ping my Mac and captured the
results into files:

        ping -i 0.064 my_ip_went_here >pingFileName

(The 64 millisecond interval matches the frame rate of Avara and ping packets
are similar in size to Avara packets.)

Here are the results:

14400, V.42bis                                  137/152/268 ms
28800, V.42bis                                  141/150/197 ms
28800, MNP4 error correction                    170/243/339 ms
28800, MNP5 compression                         182/243/316 ms
28800, no error correction or compression       148/152/159 ms

The minimum times are pretty useless for Avara and the maximum times are
warped by the first few packets (they take longer than later ones), so
the average is really what you should be looking at.

I didn't let the pings to run for more than a little while, so the averages
will vary by a few msec because of random timing changes.

The tests were done with a PowerBook 190cs, OT1.1, FreePPP2.5v2 and a
GlobalVillage PowerPort Platinum Port PC-card modem.

The most notable thing is that a 14400 connection has virtually identical
roundtrip times with the 28800 and that turning off V.42bis compression
didn't result in any change to the roundtrip times.

Another thing to note is that MNP is clearly _very_ bad for latency.
If your modem only does MNP, TURN IT OFF if you want to play Avara. If
it does V.42bis (as most modems do now), leave it on, because the extra
bandwidth you get can be very useful.

Test conditions may change the results you get. For instance, using
Mac TCP Watcher gives me 126/130/140 using 28800 and V.42bis. For
some reason, that's 20 msec faster than pinging the other way even.

The difference _IS_ significant, because a 128msec ping time allows
for an LT1 game with a T1 player or LT2 with another modem, while
150 msec requires LT2 to a T1 and LT3 with another modem. Beta testing
showed that LT2 is occasionally possible between two modem players,
but LT3 is much more common.

Latency & packet loss are the two evils of playing over the internet.
Latency is caused by modems and packet loss by the internet. ISDN or
T1 connetions have significantly less latency, so they only suffer
from the occasional packet loss (and Avara has built-in safeguards
that minimize the effects of packet loss).

Summary:

        14.4    Just as good as 28.8 for 1-1 games.
        MNP     Bad for you.

If uncertain, try the following modem init string:

        AT&F%C0\N0

It will turn off compression and error correction. If you see no improvement,
go back to what you did before.

-- 
Juri Munkki jmunkki@iki.fi              Life is easy when polygons are cheap.
http://www.iki.fi/jmunkki                  Windsurfing: Faster than the wind.