Re: POP3 Audit

[Brian Stafford <brian stafford uklinux net>]

On 2001.10.06 15:47:33 +0100 Carlos Morgado wrote:
> On Sat, Oct 06, 2001 at 03:51:13PM +0200, Ali Akcaagac wrote:
> > 
> > hmmmm... hmmmmm ?
> > 
> > yeah... well i don't want to go to my university's computer center and tell
> > them that their mailserver is crappy not to mention that they still use
> > obsolete novel network and other crappy stuff... so mentioning something
> > like this may piss them off :) but it's hard getting quality personals
> > nowadays *evilgrin*
> > 
> well, it's not so much "stupid server" as "server who don't expect a client
> to fall on it's arse in the middle of a connection".

In my book, that's a stupid server. <evil grin>  However I observe that for a
command-response protocol, a server is easier to implement than a client.  So
there is no excuse for servers to get upset just because a client drops a connection
half way through a session.  Supposing there is a power failure, or the client
is outlook and 'doze crashes etc etc.  There are lots of reasons for protocol
sessions to fail in addition to client bugs.  This is not a problem to a client -
just make a new connection.  But a server is a long lived process and MUST be
able to recove from such situations.
 
> don't go tell anyone their server is crappy and doesn't behave when you
> client breaks miserably ;)

The POP client in Balsa is not the best.  Right now, it can only be patched to
handle some common failure modes.  The best solution is to replace it entirely.
A good approach to clients is to implement them as a state machine - like a
certain other protocol client I know of <sly grin>

Brian