smofsweb@bostonbaden.com
As a "behind the scenes" backstop measure, any message that the Infobot couldn't process was saved in another mailbox for later perusal by myself. I didn't advertise this feature, but this meant that if someone sent a message to info@lacon3.worldcon.org that said "Please send me infomation on the dealers room" I would find their message later and take care of it myself.
I eventually added a feature to the Infobot where if it couldn't understand a command, and the first word of the sentence correponded to a filename, it would assume a "Send" was desired and send that file back. So if someone just wrote a message that said
dealers
it would send the file 'dealers' back, along with a header explaining the 'send' command. (Incidentally, the Infobot is still up at this writing, if you want to play around and see how it all works.) It would only look at the first line, so a message that says
dealers
art.show
hotels
would only get one file back, but the instructions enclosed with it explained how to get all of them; a message that said
send dealers
send art.show
send hotels
would get all 3.
The "Send" command (and most of the commands) have a "to" option, so in the case of someone else needing dealers information, instead of writing back directly to the person (say her e-mail address was hazel@ddb.com) I'd forward a message that said
send dealers to hazel@ddb.com
back to the infobot.
Finally, the fact that it could parse the first word of a sentence meant that I could put in files to match common mistakes. A message sent to the Infobot that said
please send me information would result in the file named 'please' being sent back! which was of some help.
The Mreply package has commands 'index' and 'help' built-in, but since they were equivalent to 'send index' and 'send help' we didn't need to emphasize a bunch of different commands when we explained how to use it. In fact, I kept it simple by telling people that the two commands they needed were 'send' and 'query' (for membership lookup), and 'send index' would get them a list of files.
I felt that I'd built a system that worked well, when I could look in the log and see entries that went something like this
12:00 pm: I want information on the art show (INVALID COMMAND)
and all this happened while I was logged off, and nobody was "on duty" at
the e-mail address at all.
I can't stress how helpful Mreply was in setting up something that WORKED
and worked well. The biggest challenge was figuring out what to send in
response to unprocessable queries (which I also had available as the file
'info') and how much to explain in the 'help' file.
The problem is balancing the need to keep it concise, with the need to
explain enough so they can find their own answers and also to answer some
common questions.
I eventually started a FAQ list. (And I encouraged people to try 'SEND
FAQ' as their command.) I think that most on-line people expect that a
FAQ can run several pages, and so won't be overwhelmed when a longer
document comes back.
Something I didn't realize at the time, but figured out late in the game,
was that the questions change over time. With a Worldcon, you can
probably expect the questions on certain major issues (Hotels, Art Show
deadlines, and so forth) to crop up about a week to a month before you
have answers ready for them.
What do you do about it?
Well, the answer is often "We're going to have information later, please
try again in a week/month."
This, in turn, led to one of the questions I myself would ask certain
Other Departments: "I *know* you don't have an answer yet. What I want to
know is, *when* will you be able to announce the answer?"
How you and your concom figure this one out is up to you guys. But it
would be good to know that this situation is going to come up, so you can
be as ready for it as possible.
12:20 pm: SEND INDEX (FILE SENT)
12:30 pm: SEND ART.SHOW (FILE SENT)
Previous | Top |
Next