I'm proposing to change the ns-3 default for Wifi to use the new NistErrorRateModel, which is validated against testbed results:
http://www.nsnam.org/~pei/80211ofdm.pdf

The patch would be:
--- a/src/helper/yans-wifi-helper.cc	Tue Jun 15 11:27:51 2010 -0700
+++ b/src/helper/yans-wifi-helper.cc	Tue Jun 15 11:37:55 2010 -0700
@@ -183,7 +183,7 @@
 YansWifiPhyHelper::Default (void)
 {
   YansWifiPhyHelper helper;
-  helper.SetErrorRateModel ("ns3::YansErrorRateModel");
+  helper.SetErrorRateModel ("ns3::NistErrorRateModel");
   return helper;
 }

Note that this patch causes these four regression tests to fail, presumably because the Yans model is more optimistic, but this must be confirmed:
FAIL: TestSuite devices-mesh-flame-regression
FAIL: TestSuite devices-mesh-dot11s-regression
FAIL: TestSuite routing-aodv-regression
FAIL: TestSuite routing-olsr-regression

Another option to shift usage to this error rate model would be to keep YansWifiHelper untouched, and create a new NistWifiHelper that reused Yans components, and deprecate the usage of YansWifiHelper in the examples/tutorials.  Since "Nist" is limited to the error rate model, it seemed that the above patch might be more straightforward, however.

Any other opinions?