On 01/25/2013 06:43 PM, Phil Pennock wrote:
> On 2013-01-25 at 11:59 +0000, Jeremy Harris wrote:
>> There's a slight conceptual issue in that the order for the ACLs being
>> run would not match the order of the SMTP protocol-visible actions.
>
> How so? The DATA ACL is triggered by the "." sent by the client.
> Earlier, there's a pre-data immediately after the DATA command itself.
Point. I was only considering the response to the dot, not the dot itself.
PRDR responses, when done in full, split the two apart.
The proposed alternate ordering of PRDR vs. DATA ACL calls
(existing/main DATA ACL first) would mean one of
- The DATA acl has to run all possible data-scanners and configurations
thereof, even if none of the recipients care about a particular one
- the DATA acl has to walk the recipients list and build a merged set of
scans to run, then run them
- the RCPT acl has to build the merged set, and the DATA acl run them
- (as current) the PRDR acl runs the set particular to each recipient
(possibly taking advantage of cached results from previous recipients)
I don't see a major advantage in the proposal. Possibly the case
where there's a single scanner which returns a score, and recipients
have different tolerance levels but all care to some extent (this would
use the first implementation style).
>
>> I have some OCSP stuff waiting in the wings it'd be nice to also get
>> into the next release.
>
> Is it on a publicly visible branch?
Not quite. Hoping for within a few days.
--
Cheers,
Jeremy
