Is comint the way to go?

[Mathias Dahl <brakjoller@gmail.com>]

 I want to write a command line interface to a program that isn't
command line driven.  In fact, it is not even a program, it is a
database with a specific schema.  What I want to do is create a
package that let's the user navigate this database using familiar
command line commands like `cd' and `ls'.

 When the user does `cd', what happens in the background is that one
or more SQL-queries are sent to Oracle's Sql*Plus tool, and I massage
the response into something that fits nice on the command line.

It would look something like this:

 myapp> cd Implementation
 myapp> ls
packagefoo
packagebar
...
 myapp> cd packagefoo
 myapp> ls
file1
file2

 None of the above is "real". Each command executed by the user
actually executes a specific query, depending on the context
(basically depending on where the user has previously "moved" using
`cd'), and the output will look like the output from working with a
file system.

 My question is, is comint the right thing to use here? Or, must
comint derived packages always be built around a command line app? Am
I better off doing this more or less from scratch (I have working code
to make my queries, so that's not the issue, the issue is the user
interface)?