But how does this handle the multiple-address case? Or even the
DHCP-based DSL user?
Once the email address has been typed, nice sensible defaults appear
in the rest of the dialog. The user may change settings at will, or
just click OK.
I guess I'm not disagreeing that SRV won't help, but in a practical
sense I'm not convinced it really solves much of a problem.
A few years ago I wrote a linux installer which generally supplied
correct network details. People hardly ever needed to type in their
Was the need to type in a netmask a big problem? No way. Did people
love it when the problem went away? Yes they did.
There's a big difference between a UI that provides a default but
prompts you with an opportunity to change it, and one which silently
provides a default and if it doesn't work you have to go hunting for it.
In the first case, the expert will appreciate just being able to accept
the default, and the naive user will appreciate that it "just works"
(some of the time). (The naive user doesn't understand when something
"just works" because of good design vs. when he just got lucky)
In the second case, the naive user's reaction is almost the same.
the difference is that when the naive user consults an expert and
the expert says (e.g.) "what netmask did you use?" the naive user
won't have any idea what the expert is talking about.
When an expert tries to use the second type of UI he will either be
annoyed that he has to chase down how to specify something that he
knows needs to be specified (= cannot in general be inferred), and/or
he'll be annoyed that when it doesn't work and he has to look for
whatever hidden knob might be causing the software to break.
FWIW, I don't think it's appropriate to use SRV as a way to do
hidden defaults - you want a late binding of domain to
POP/IMAP/SUBMISSION server (so that the admin of that domain
can change it on short notice), not a binding that is made when the
user configures the software and which defaults to where the SRV
Power corrupts; Powerpoint corrupts absolutely.