For the record, in case this patch is missed at the end of a very
long message.
> Of course we could introduce a hack where a special lookup result
> of "NONE" (or some other magic string) means don't use a relayhost.
>
> Postfix would then look like this:
>
> /etc/postfix/main.cf:
> relayhost =
> sender_dependent_relayhost_maps =
> hash:/etc/postfix/sender_relay static:[my.isp.com]
>
> /etc/postfix/sender_relay:
> [EMAIL PROTECTED] none
> [EMAIL PROTECTED] none
>
> which is much cleaner.
>
> Introducing "NONE" (or whatever) requires a source code change.
> It also will raise expectations: people will try to use it in other
> tables and be disappointed that it does not work there, just like
> you tried to use "smtp:" in a place where Postfix expects "host"
> or "host:port".
There is prior art in Postfix with magic lookup results.
For example, header_checks allows you to search a list of tables.
When a table returns a result of DUNNO, Postfix stops searching
and behaves as if the result is "not found".
In my last example above, support for DUNNO (instead of "none")
would also solve the problem. There is no need to introduce a new
magic lookup result, just one line of code that works in the same
way as people already know from elsewhere in Postfix.
Wietse
*** ./src/trivial-rewrite/resolve.c- Wed Dec 12 19:23:26 2007
--- ./src/trivial-rewrite/resolve.c Fri Oct 3 11:20:22 2008
***************
*** 511,517 ****
&& (relay = mail_addr_find(rp->snd_relay_info, *sender ?
sender : var_null_relay_maps_key,
(char **) 0)) != 0)
! vstring_strcpy(nexthop, relay);
else if (*RES_PARAM_VALUE(rp->relayhost))
vstring_strcpy(nexthop, RES_PARAM_VALUE(rp->relayhost));
else
--- 511,518 ----
&& (relay = mail_addr_find(rp->snd_relay_info, *sender ?
sender : var_null_relay_maps_key,
(char **) 0)) != 0)
! vstring_strcpy(nexthop, strcasecmp(relay, "DUNNO") == 0 ?
! rcpt_domain : relay);
else if (*RES_PARAM_VALUE(rp->relayhost))
vstring_strcpy(nexthop, RES_PARAM_VALUE(rp->relayhost));
else
