A DOM parser builds a tree of nodes that represent the XML nodes, elements and attributes that comprise a document. It's convenient and intuitive to use, but memory intensive and slow.
A SAX parser is "event driven." You write a handler that is notified when the parser sees an opening or closing element, an attribute, etc. You then do whatever you want with that information. It can be hard for some people to understand, but it's fast and memory-efficient.
You cannot modify an XML document file using SAX you can only read it. You cannot create a new XML document from scratch using SAX.
Close but not exactly true - You can use SAX to modify an input XML document if you write the modified document on the fly. Since you only get access to one element at a time, you are limited as to the types of modifications BUT there is no limit on document size the way DOM limits you to a document that fits in memory. The Cocoon project does some amazing things with "pipelines" of SAX events. Bill