> > I would think this stat would be difficult to come by since a lot of 
> > smtp servers don't return a name or version in the smtp header for 
> > security reasons.
> 
> Yah, I just imagined someone would have done a study.  I've seen numbers like
> 70% sendmail but don't know how accurate that is.

Someone has, but it's not publicly available. 
The 70% is not accurate.
Despite being out of date, the numbers presented at 
	http://www.ohse.de/uwe/surveys/smtpauto.html
show a trend worth paying attention to. 
Also note the sampling method is reasonable. 

Andy