Re: Announcing bug tracker

[W B Hacker <wbh@xxxxxxxxxxxxx>]

Simon 'corecode' Schubert wrote:

*SNIP*

> No, we can't go this way.  If a mail arrives at an invalid address, it 
> will be bounced.

Symantics.

Hopefully what you *actually* do is not exactly what you said.

Your MTA should be responding with an smtp error code *during* the connection, 
and these cannot reach anyone but the still-attached distant MTA, hence harmless 
as well as RFC-mandated.

IF that is a 'proper' MTA, IT will generate a descriptive message back to its 
own user (and no other). A zombie will simply ignore it all.
Both are harmless to others.

A 'bounce', OTOH, is a separate message your MTA creates and sends *after* the 
connection has been dropped, and is addressed to the purported source, which may 
be forged (though that is detectable and avoidable..).

Such traffic may either reach an innocent bystander, OR in turn be bounced back 
to you - rewind.

That's where the possibillty of collateral damage arises - not from anything you 
reject with an smtp response during the life of the connection (prior to DATA 
phase anyway).

> So what does it break, then?  I could see your point, but I don't accept 
> it without proper reasoning.

If these are being handled 'downstream' of the serving MTA, such as in an MUA, 
getmail, fetchmail script/filter, then they *are* a problem, as these are 
time-displaced from the ability to exchange peer MTA connect-time handshakes.

Hopefully clear now.

Bill