ietf-mta-filters
[Top] [All Lists]

Re: List of open issues with Sieve reject draft (draft-ietf-sieve-refuse-reject-02.txt)

2006-07-11 08:33:07

On Mon, Jul 10, 2006 at 12:02:25AM +0200, Arnt Gulbrandsen wrote:

As I read the current draft, such an interpreter is not allowed to 
reject the RCPT TO command, because it does not yet know whether the 
script is valid: The script may also try to execute some action which 
is incompatible with reject.

This looks like that old question: does the prohibition on
"reject+redirect" (or any other combination of actions defined as
incompatible) produce an invalid script, or an invalid action?
Per RFC3028 this is an implementation choice.  With the latter
choice, a script:

    reject "go away";           # is executed
    redirect "mem(_at_)somewhere";   # gives an execution error

and if they are reversed, the redirect is executed but not the reject.
If an implementation chooses the former (wait until the end to execute
actions, doing none of them if there are conflicts) then you might have
to address the question you raise, but I don't think that should be in
the spec.


As for reject + fileinto: the one thing I could see it useful for has
already been mentioned, and that is administrative logging in order to
be able to debug script design decisions.  It's tempting to just say
"solve that problem another way, outside of sieve" but I can certainly
see how using fileinto is convenient for this.  I suppose an
implementation could have an vendor-specific capability to enable this,
not condoned by the spec.

Yours,
mm