On Tue, 02 Aug 2011 16:43:13 +0200 Stephan Bosch wrote: > On 7/29/2011 8:53 AM, Nikita Koshikov wrote: > > On Thu, 28 Jul 2011 20:43:45 +0200 > > Stephan Bosch wrote: > > > >> Would the header name need to be configurable, or is > >> X-Sieve-Redirected-From good for all purposes? Regards, Stephan. > > I saw this > > > > X-ResentFrom: > > X-SRS-Rewrite: > > X-Forwarded-(To/For): > > > > So the name is not significant. If it's not so difficult to code - than > > configurable header of course is better. But it's only a matter of taste. > > Adding more settings is best to be avoided. I've added the header: > > http://hg.rename-it.nl/dovecot-2.0-pigeonhole/rev/81b37c0055c3 > http://hg.rename-it.nl/dovecot-2.0-pigeonhole/rev/efad75f779de (d'oh) > > > One thing to note for implementing: > > It will be much more simple for setting up srs, if sieve will export to > > environment some variables indicating need for rewrite. And than we will be > > able to write wrapper for sendmail_path that will selectively choose what > > should we do: > > For example, I don't want srs to happen if user is forwarding for local > > domain mailbox, or if action that trigger mail sending is - "out of office > > reply". > > So, wrapper will analyze need srs or no, and after all invoke srs binary > > for signing, or just call pure mta. > > I am not sure Sieve has access to enough information to make such an > assessment. And I don't think adding SPF/SRS-specific code to Sieve is a > good idea. It would also not be very straight-forward since the > low-level mail sending code is part of Dovecot itself. > > Regards, > > Stephan.
I meant export to environment some variables, like dovecot does: USER= IP= MAIL= .... Adding more like: SIEVE_ACTION= SIEVE_RECIPIENT= etc And then execute sendmail_path. Thus program\wrapper can be aware of mail originated from sieve. This is not specific spf\srs code. Seemed, that this section of code need to be in dovecot-lda region, so maybe Timo will be able to catch it. Thanks for your work.