[Q] Databases -- Gadfly vs Metakit vs SQLite

[achrist at easystreet.com]

I'm looking for a simple database for a single-user Windows desktop
application.  The user's data will be text, maybe 10-20 data tables,
with identifying keys and a few short fields plus plenty of long
text memo-type fields.  I should like to plan for databases up to
about 50k rows and 100 MB of data, but the typical will be far less
than this.

It looks like Python gives me 3 good choices for this, Gadfly,
SQLite, Metakit.  I'd like comments that compare and contrast these
for such an application.

This is a desktop productivity app, so I'd like to get snappy response,
no 1 or 2 second delays for a single update or retrieval.  This should
work to such a standard on a user's 300 MHz, 64 MB Windows machine.  

Does Gadfly need to fetch the entire database into RAM?  This would be
fine for me 99% of the time, but I don't want to worry about the user
with the big database and the small machine.

Other criteria:

1. Very high reliability -- Which of these can give us zero database
failures from 1000's of users?

2. Simple -- We want to hide all details of the database from the
end-users, and easy to program wouldn't hurt either.  It looks like
all three will meet this criteria.  Any traps?

3. Storage efficiency -- We don't want 10 MB of user data to take 
100 MB of disk space.  I know that disk is cheap, but we want the
data to be easy to back up, send,  and transport.

The hierarchical non-SQL aspects of Metakit are probably OK.  

Can anyone with experience with more than one of these advise?

Thanks much in advance.


Al