Dan Weber <[EMAIL PROTECTED]> said:

[snip] ... >> we still have to find a
>> way of having the daemons use the changed config (and at not bombard
>> the server holding the config with a gazillion of queries every
>> second).
>>
> Caching -- check every 600 seconds and that will allow a small grace period.

Depends how instantaneous we expect the changes to be. Would requiring a
HUP be too much? Perhaps, because if we were on the box to give a HUP then
we'd just copy the file anyways. However, if changes take five minutes to
propagate, a fatal config error would be fatal for that long.

Everyone put on your thinking caps, we need to figure this out!

>> For sure, it's not the UNIX-way. It should be possible to make the
>> setup using the config-file for simpler setups.
>>
> It has some advantages if you want to be able to manage many dbmail servers
> at one given point, tho for most people none is gained.

The database is required for DBMail to work anyways, so there's no hard in
requiring a database for the configuration ;-) Because it scales
seamlessly from one DBMail box and a local database to a thousand DBMail
boxes and replicated databases, it is an excellent option. Being able to
quickly scale an email system from "does this work?" to "will this make 1
million or 2 million dollars this year?" is critical for lots of business
environments.

I do think that there needs to be an algorithm for config file overrides,
to be used for doing something different on a testing server in a cluster
and in the degenerate case, to have all options in the config file for a
newbie.

Aaron

--

Reply via email to