Excerpts from mail: 22-Nov-94 Re: news server in AFS? "Perry E.
Metzger"@imsi. (2251)
> > But that's the same as having the delivery agent put it directly in your
> > inbox. The only difference is that you have to kick it off manually instead
> > of having it happen automatically.
> Remember that in some sense you can say that the mail ALWAYS
> ultimately ends up in the user's directory. There is, however, good
> reason to have the user kick off the act of moving the mail instead of
> having this "difference" of having the mail system do it.
Excerpts from mail: 22-Nov-94 Re: news server in AFS? "Perry E.
Metzger"@imsi. (2251)
> Especially when you try to run high reliability environments, its nice
> to have single points of failure cause as little disruption as
> possible. To me, POP works very nicely in this sort of world -- it
> scales perfectly, its easy to secure, its easy to manage, etc. Random
> hacks with Sendmail and AFS make me far less happy overall and they
> don't seem to have any obvious benefits other than making users
> operating AFS over slow lines from their home not need to run a single
> program to load their mail on the main campus network before trying to
> read it.
Discussions similar to this one on mail delivery have appeared on this
group before. I don't recall any previous discussion of mail
*notification*, however. How does the user know to read his mail?
Around here, our non-AMDS users tend to use "xbiff", which requires one
connection and mail server process per user. Does this "scale
perfectly"?
AMDS, which unfortunately doesn't appear to have a future, delivers
e-mail into an AFS users home directory. The "console" program loops
waiting for this directory to change. As long as it doesn't, these
status checks generate no network traffic. As soon as a message
arrives, the server for the user's mail directory breaks the callback on
that directory, "console" sees the additional mail messages, and updates
its display of the number of unread messages. Pretty slick compared to
"xbiff", I'd say.
Keith Gorlen
Division of Computer Research and Technology
National Institutes of Health
Building 12A, Room 2033
12 SOUTH DR MSC 5624
BETHESDA MD 20892-5624
Phone: (301) 496-1111, FAX: (301) 402-2867
Internet: [EMAIL PROTECTED]
