.ds em "\~\[em] \" note trailing space
.ft TI
.ps 14
.vs 16
.in 2i
.po 0
.ll 21c-2i
.fi
.ad l
.sp 1.5i
.ds lqtext "\&\f[TBI]\%lq\[hy]text\fP
.nh
\f[TBI]lq-text\fP
uses an inverted index which is generally a little larger than half the
size of the original data.
This compactness is achieved by careful control of internal `fragmentation',
and also by proprietary compression techniques which are designed
to avoid comprimising speed.
The individual text files can be compressed\*[em]perhaps using
the Unix\[tm] `compress' utility\*[em]or archived, as desired,
because \*[lqtext]
has all of the information it needs in the index, and does not need to
refer to the actual files, even for matching complex phrases.

You can have multiple \*[lqtext] databases,
specified by an environment variable, which can be over-ridden
at any time with a command-line option.

You can specify a `document path', which is a list of directories separated
by colons.
When \*[lqtext] searches for a file whose name doesn't start with a
\&\s-2\fB/\fP\s0, it looks in each of the document path directories in turn.
This means that you can retain flexibilty to move files around after
they have been indexed, as well as slightly reducing the size of the database.

A configuration file in each database lets you specify a file of common
words to exclude from the index, and also a default document path.
This is an \&\s-2ASCII\s0\& file, and can be created with any text file;
it can include comments.

There is a screen-based front end, although all of the functionality
is also available through the Unix command line.
