Good idea?

Hello group.

I have an enormous list of user-identities in my application, some of our
clients even have more than 100.000 entries. Until so far, we used to store
these identities in a section-based, ini-lookalike file, like this:

[Username1]
realname="John Doe"
enabled=yes
[EMAIL PROTECTED]
address=street
etc=ectetera
next=nextKey
key=Value
blah=blahblah

[Username2]
realname="Jane Doe"
enabled=no
[EMAIL PROTECTED]
address=street
etc=ectetera
next=nextKey
key=Value
blah=blahblah

etc...

with some 15 to 20 different keys for each user.

Obviously, this approach is very time-consuming when one wants to find any
user where  e.g. his email is in the somewhere.com domain (as in the
example) or e.g. any user that has enabled=no
Now we have to read all sections and then read all the email-keys.

Is it a good idea to store such data in a berkeley db, or should we consider
something like MySQL?

Gr, Hans