IMO the major challenge will be to extract the appropriate information from each feed in semantic form. Wikipedia describes mashups as:
There are many types of mashups, such
as consumer mashups, data mashups, and
enterprise mashups. The most common
type of mashup is the consumer mashup,
aimed at the general public.
Data mashups combine similar types of
media and information from multiple
sources into a single representation.
One example is AlertMap, which
combines data from over 200 sources
related to severe weather conditions,
biohazard threats, and seismic
information, and displays them on a
map of the world; another is Chicago
Crime Map, which indicates the crime
rate and location of crime in Chicago.
The classic mashup - Chicago crime - works because key information such as dates and geolocations are available semantically. Other types of common information are persons, organisations, and domain-specific identifiers.
When you have identified these you may wish to consider the RDF-based tools that the semantic web is developing. Note that governments are starting to emit their data in RDF so I would see this as a key technology
If your web pages do not have semantic information immediately you will probably have to create screen scrapers and HTML parsers. That s not very glamorous, there are no special tools and tends to be just hard work.