Tony Hansen wrote:
Nothing is broken. Sendmail can be invoked using "sendmail addresses"
just as easily it can be invoked using "sendmail -t". When your program
that uses sendmail is upgraded to recognize the NoReply fields, continue
to use "sendmail -t" for the To/Cc/Bcc recipients, and then use
"sendmail addresses" for sending the message to the NoReply recipients.
I don't get it. It seems like the only reason to use "-t" is if the
sending program doesn't want to bother parsing the header. once you go
to the trouble of modifying it to parse NoReply fields, why not have it
parse all recipient fields and then do "sendmail all-addresses..."?
I hope "sendmail -t" is NEVER upgraded to support NoReply headers. That
would cause the worst of all possible worlds. As a "sendmail -t" user,
you'd never know if your invocation of "sendmail -t" included those
NoReply headers or not. Say your "sendmail -t" doesn't support NoReply
and you write you code to invoke "sendmail addresses" to handle NoReply
headers. Then some helpful admin upgrades your "sendmail -t" to support
them. Suddenly the NoReply recipients will be receiving TWO copies of
the messages.
another reason why I wouldn't rewrite the submission code to send
NoReply fields separately.
If sendmail WERE to be upgraded to support NoReply headers, then it
should be done using a DIFFERENT option. (For the version of sendmail on
my system, it looks like -T might be available.)
makes sense to me ... especially if -T causes an error with existing
"sendmail" and clones. (on my system sendmail -T complains about an
invalid date, so maybe it does something undocumented.)
Keith