Udi Manber -- Software
The most prominent software packages that we developed are
The Search Broker
The Search Broker
provides a two-level web search paradigm by forwarding
each query to a specific search engine.
Users specify the subject of their queries as the first
word in the query, and the Search Broker automatically
reformats and forwards their query to a relevant search
engine selected from a list of over 400 topical search engines.
For example, try
Glimpse is a software package for searching file systems.
We have distributed it since 1993, and it is now at version 4.1.
Glimpse provides a wide range of options (e.g., search for approximate
patterns, regular expressions, filtering by file names and content)
and it is designed both for small personal file systems
and large servers. It uses some of the algorithms we originally
developed for agrep and some new ones.
It has been ported to most UNIX systems, and has been licensed to
over 50 companies.
For example, the Virtual Hospital
is fielding more than 100,000 medical queries a day using Glimpse.
Harvest is an integrated system to collect, extract, organize, index,
search, cache, and replicate information across the Internet.
Besides being a very useful system, already deployed by hundreds
of sites (including, for example, the CIA, NASA,
the National Academy of Sciences,
and the Government Printing Office),
Harvest serves as a research vehicle, generating
new problems and ideas.
Netscape's Catalog Server is a commercial version of Harvest.
Network Appliances's cache is a commercial version of the
WebGlimpse is a search tool
that provides a flexible combination of browsing and searching.
WebGlimpse automatically attaches a
small search box to the bottom of HTML pages, and allows the search to
cover the neighborhood of that page.
Neighborhoods can be defined in a variety of ways.
They can even include remote pages.
WebGlimpse has been distributed since June 1996;
it is currently at version 1.5.
is a collection of tools to use glimpse in HTTP servers.
In addition to the usual features of glimpse, glimpseHTTP
integrates search with browsing. If the data resides in a large
hierarchical tree of directories, glimpseHTTP indexes all of them at
once, but keeps their structure so that when a user browses through them
and decides to search from a certain place, only data at or below that
directory is searched.
GlimpseHTTP also translates results to HTML and provides several
other facilities to make search more convenient.
More than 1000 WWW sites have used it at one time or
The US department of Justice,
US Department of Health and Human Services,
The American Red Cross ,
NASA Marshall Space Flight Center
and more than 150 universities.
(GlimpseHTTP is no longer maintained. It has been superceded by
agrep is an improved grep that allows approximate matching,
Boolean queries, user-defined records, and a whole lot more.
Back to Udi Manber's Home Page