getting off this mailing list

[Ed Cogburn]

Stephen R Laniel wrote:

> On Sun, Jul 03, 2005 at 03:50:44PM -0400, Ed Cogburn wrote:
>> Given that fetchmail is nothing more than a shallow wrapper around the
>> POP3 internet mail transfer protocol
> 
> Well ... not really. Look at it instead as a tool for
> accomplishing one task: getting mail from a remote machine
> to your own machine. Make that tool work
> 
> 1) reliably (messages aren't deleted off the remote machine
> until they've been committed to persistent storage on the
> local machine),
> 
> 2) using any mail transport protocol (all variants of POP,
> IMAP, etc.),
> 
> 3) using any authentication and encryption mechanism (SSL,
> SSH)


The GNU folks have the "mailutils" package which is trying to provide this
as a common service, but I believe that is still a work-in-progress, and
the Desktop Environments couldn't wait for a standard to emerge, they
needed flexible POP3 support now/then.  This plus the fact that POP3 itself
is a basic simple protocol that can easily be reimplemented "in place"
wherever else its needed lead to many mail readers supporting it directly. 
If a flexible, easy-to-use POP3/IMAP daemon establishes itself as a
standard cross-platform method of mail interchange, I expect most clients
of POP3 services will eventually move to that.  Unfortunately, fetchmail
isn't that flexible, easy-to-use daemon that people want, so its still
*easier* to just implement POP3 directly rather than craft an interface to
fetchmail.

In the case of the DEs, they implement it once as a service of their
environment in a way more flexible for them, i.e. KDE and kmail can do the
3 things you mention above, and they do it that way for the same reason why
neither KDE or GNOME provide you a GUI frontend to ls, they needed to go
beyond what ls could do for them, and since it was relatively easy, they
did so.  Note that your #3 is really a separate protocol/feature that isn't
just used for mail, so shouldn't be implement within mail services, but
rather implemented separate with the mail services using it as a client
(fetchmail itself is using libssl for this).  If we ever get a flexible
standardized "libnetmail" (or "netmaild", a daemon designed for other
programs to be its clients, ala inetd) to handle these things, then what
you're suggesting will likely happen.