[Nut-upsdev] Starting drivers without entry in 'ups.conf'

[Arjen de Korte]

>> Until the release of nut-2.0.5 it was possible to have a working
>> configuration without configuring anything in 'ups.conf'. Since then,
>> this no longer works, as the server will only look for driver sockets
>> for UPSes that have been configured in 'ups.conf' (since the socket is
>> named after the UPSname).
>>
>> Drivers will use the old naming scheme '<driver>-<port>' if the UPSname
>> can't be found in 'ups.conf'. From time to time, people are confused by
>> this, especially since this used to work up to (and including)
>> nut-2.0.4. Some distributions even offered this as the standard way of
>> configuring single UPS systems.
>>
>> Now that this is no longer possible, I think it is better to refuse to
>> start in case the UPSname can't be found, to prevent problems later on
>> in the configuration process. Any thoughts on this?
>
> since this behavior has been kept for debug purpose, this suits me fine.
> It will avoid many users confusion, and is not a big loss for us, the
> developers.
> Moreover, this will force packagers to update their legacy scripts.
>
> So, you can apply your change to both the trunk and the Testing branch.

Then I have a couple more questions. Since this also means that UPSes must
be configured in 'ups.conf', do we still keep the option to configure a
port on the commandline for starting a driver?

Currently, the syntax for starting a driver is

    <drivername> [OPTIONS] [<device>]

Since we now require that drivers are configured in 'ups.conf', I think we
could probably leave out the optional <device> here. So my proposal is to
mandate the use of the '-a id' option (like discussed already) and remove
the optional specification of the <device> (which would override the
'port' from 'ups.conf'). The proposed syntax would then be

    <drivername> -a <id> [OPTIONS]

Any thoughts?

Best regards, Arjen