Looking over the models runs:

* Both CMC (00Z) and UKMET (00Z) match the NHC westerly track pretty closely.
* NOGAPS (00Z) takes the system more west northwest.
* GFS (06Z) and GFDL (00Z) are swinging the system southwest around the tip of FL.

I find the last two curious. While the GFDL has been consistantly showing a southward swing, it is the 00Z run. The GFS, which had been showing the system sticking to the Atlantic yesterday, now, on the 06Z run, is do a sudden drop south then southwest just as Katrina nears the coast. It looks like both the GFDL and GFS actually cause the system to be disrupted slightly near the coast, but then reform it slightly south allowing the jog around FL. If this pans out, it "might" be good new for east coast of FL, but certainly not for the gulf.

Given the past reliability of the GFS and GFDL models, we can't ignore these potential tracks.

Model tracks: http://moe.met.fsu.edu/tcgengifs/
I used 850mb Vorticity for the tracks, except GFDL where the 850mb is missing. There I used the 310K Potential Vorticity.