Rule based vs. Signature based detection engine

["Tom Fulton" <tfulton9909 () comcast net>]

The Snort 2.0 book by Jay Beale, et. al., (p. 142) explains that Snort is
rules-based, "which collects and correlates packets based on rules" and that
this is better than a signature engine which is nothing more than a
"definition of an attack".  Can anyone expand on this clarification?  I'm
under the impression that a rule in a *.rules file is basically a
"signature".  Do you think Jay is referring to the ability to have
pre-processor plug-ins that can normalize data before running against the
signatures (sorry, I mean rules)?  Aren't they basically the same thing when
it comes right down to it?