Things needed (was: Re: Do we really need Korn ?)

[Scott Wheeler]

On Sunday 28 April 2002 02:22 pm, Andy Fawcett wrote:

> I'll look at it sometime. Maybe they should combine, and use the best 
> features of both, and also provide better hooks to kmail?

Well, if you poke through the Korn source, you'll see that it's already using 
some of the KBiff code.  I think that KBiff has a much better interface, but 
it has its problems too.  However, I haven't seen any problems that KBiff has 
that Korn doesn't share (though I haven't used Korn as much).

So, specific issues with KBiff (that I think are shared by Korn):  KBiff does 
it's own implementation of the various mail protocols rather than using IO 
slaves.  Also it uses system libraries for socket IO instead of QSocket and 
friends, and despite the option for async IO in the configuration, it doesn't 
seem to use it.  This is particularly annoying when the event loop is blocked 
waiting for a socket to time out (if you for instance unplug your ethernet 
cable, you may have to wait over a minute to be able to use the interface 
again).  The independent socket implementation also made it non-trivial a 
while back to say add SPOP support (which I would *really* like to have in 
something that sends my password over the internet every 5 minutes).

Also, KBiff currently uses 'kmail --check' as its default way of checking the 
mail.  I've changed that here to try DCOP calls first and then only if they 
fail to launch KMail, though it should be implemented in the code rather than 
in my command line edit.  I don't think most people are going to put '( dcop 
kmail KMailIface openReader && dcop kmail KMailIface checkMail ) || kmail 
--check' in there.  

So, despite all of this, I think KBiff is great; but I think both Korn and 
KBiff could use some work.  KBiff has the advantage presently of not really 
having this demanded because it's not in the distribution.

Anyway, enough rambling --

-- Scott