Re: Re: config

[Gordon JC Pearce]

On Wed, 2003-12-24 at 18:38, John B wrote:

> My own view on this is that if a device is able to retain it's
configuration
data (e.g. PC app, embedded device using EEPROM etc.) then it should do.
This,
as you rightly say, means it only needs the config manager running when
it's
first configured, and if you ever want to reconfigure it again in the
future.
This is the approach I have taken with my .NET xPL framework.

This is the route I am taking with my Linux libraries, based the Linux
development package

> However, if a device cannot retain configuration data, e.g. due to
lack of
EEPROM or other suitable storage, then I can't see another way round things
other than having a configuration manager running.

In which case, though, it *must* have some unique identifier, otherwise
there would be serious problems when you powered up more than one
identical device. This might take the form of a unique serial number
burned into the device. The approach I'm working on for the Linux
development libraries is to generate a unique instance based on the IP
address and port of that particular client. For instance, if DHCP hands
the device 192.168.0.22, and it connects to port 50003 on the xPL hub,
it will call itself something like device022003.

Gordon.