David Burrowes' Blog

I'm doing some design work for the OpenSolaris Network Automagic project ( http://www.opensolaris.org/os/project/nwam/ ), which I think will end up on that site next week.

Some of what I'm looking at is how can we automatically configure network settings on a computer.

In the process, though, I've begun to wonder about some of the settings for my applications. For one prominent example, in my email apps I usually have to type in a bunch of information to set up the account: my email address, my incoming server, whether to use SSL, my outgoing server, what userid and password to use for that and so on.

Why do we have to specify all this information? Surely 99% of the users specify the same settings except for their identifier. Shouldn't the rest be able to be automatically configured?

So, the usage scenario might be like this: (1) Start new email client (2) It asks me what my email provider's identifier is, and I type in "email.frozboz.net". (3) The email client asks the server what email protocol to use (POP or IMAP or whatever), the name of the incoming server, the outgoing server, the authentication modes, the ports. (4) The email client stores these and then asks for my username and password to access my email.

You can get more comprehensive than just solving the problem for email, but I leave that as an exercise for the reader.

Really, my point is: at this point in our computing history, asking someone to do more configuration than typing in a name and password is akin to cruel and unusual punishment. I suspect that many things can be automated better than they are, and the more you can automate in this regard, the happier your users can be.

Now someone just needs to set up an email automatic configuration (Call it DACP for Dynamic Application Configuration Protocol) and get it into Thunderbird and some mail servers and then watch it gradually get adopted.