Sam Varshavchik <[EMAIL PROTECTED]> writes:
> [ ... ]
>
> This is the code in submit2.C that generates the queue ID:
>
> …
> << COMCTLFILE_MSGID << ino_buf << '.' << basemsgid << endl;
> …
>
> ino_buf is the control file's inode number. basemsgid is generated a
> few lines earlier, like this:
>
> [ ... etc. ... ]
Thanks again, Sam.
OK. Now I have changed my unique ID patch so that the id that gets put
into the "Received" header is the queue ID, as in the 3rd line of this
example:
Received: from bsd4.nyct.net (mail-out.nyct.net [::ffff:216.139.141.2])
(IDENT: root, TLS: TLSv1/SSLv3,256bits,AES256-SHA)
id 0021CB0D.40303AA5.00008701
by asfast.net with esmtp; Sun, 15 Feb 2004 22:36:05 -0500
I've attached the patch, below.
I'd like to request that this patch be put into an upcoming Courier
release (after you're convinced that it actually works, of course). I
believe that this would be a beneficial thing to do for the following
reasons:
1. This "id" field is specified as being valid for the "Received"
header in one or more RFC's, so we are not deviating from any
standards.
2. One or more other MTA's also make use of this "id" field in the
"Received" header. For example, in exim, this field appears, and it
also represents the mail queue ID. Therefore, its use is consistent
with that of other existing email software.
3. It's a minor change: only one extra line in the "Received" header,
an infinitessimal impact on execution speed, and no change
whatsoever in the message processing flow.
4. With this unique ID in the "Received" header, it is now possible to
keep state information for the life of each message. By means of
this, it's possible for events that occur during global filtering to
trigger changes to the message during subsequent local filtering.
One version of this procedure is described in the "Modest Proposal"
that I posted several days ago.
5. Nothing about this minor change will prevent other, more ambitious
changes in message processing to be implemented in the future, if
desired.
So Sam ... do you consider this to be a worthwhile change to put into
an upcoming release?
*** courier/submit.h.orig Sun Feb 15 21:03:57 2004
--- courier/submit.h Sun Feb 15 21:39:00 2004
***************
*** 117,122 ****
--- 117,123 ----
SubmitFile();
~SubmitFile();
+ char* QueueID();
void SendingModule(const char *p) {sending_module=p;}
void Sender(const char *, const char *, const char *, char);
int ChkRecipient(const char *);
*** courier/submit.C.orig Mon Dec 15 20:51:29 2003
--- courier/submit.C Sun Feb 15 22:17:45 2004
***************
*** 1088,1093 ****
--- 1088,1097 ----
line += ')';
}
+ // Add unique id here.
+ line += "\n id ";
+ line += my_rcptinfo.submitfile.QueueID();
+
line += "\n by ";
line += config_me();
*** courier/submit2.C.orig Sun Feb 15 21:26:35 2004
--- courier/submit2.C Sun Feb 15 22:33:25 2004
***************
*** 434,439 ****
--- 434,500 ----
) ;
}
+ char *SubmitFile::QueueID()
+ {
+ struct stat stat_buf;
+ char ino_buf[sizeof(ino_t)*2+1];
+ char time_buf[sizeof(time_t)*2+1];
+ char pid_buf[sizeof(time_t)*2+1];
+ static char result[sizeof(ino_buf)+sizeof(time_buf)+sizeof(pid_buf)];
+
+ char* f = strdup(name1stctlfile());
+ if (f == NULL)
+ {
+ clog_msg_errno();
+ }
+ int rc = stat(f, &stat_buf);
+ if (rc != 0)
+ {
+ free(f);
+ clog_msg_errno();
+ }
+ // This code assumes that the result of name1stcltfile() is
+ // of the form SOMETHING/tttttt.ppppp.hhhhhh, where
+ // SOMETHING is arbitrary (as far as this routine is concerned),
+ // where tttttt is the modification time of the control file,
+ // where pppppp is the pid of the process which created
+ // the control file, and where hhhhhh can contain zero or
+ // more dots.
+ char* cp = f;
+ char* slash = f;
+ for (; *cp != '\0'; cp++) {
+ if (*cp == '/')
+ {
+ slash = cp;
+ }
+ }
+ cp = slash + 1;
+ char* dot = cp;
+ for (; *dot != '.'; dot++) { /* do nothing */ }
+ *dot++ = '\0';
+ time_t tm = (time_t) atoi(cp);
+ cp = dot;
+ for (; *dot != '.'; dot++) { /* do nothing */ }
+ *dot++ = '\0';
+ int pid = atoi(cp);
+ free(f);
+ libmail_strh_ino_t(stat_buf.st_ino, ino_buf);
+ libmail_strh_time_t(tm, time_buf);
+ libmail_strh_pid_t(pid, pid_buf);
+
+ strcat(
+ strcat(
+ strcat(
+ strcat(
+ strcpy(result, ino_buf)
+ , ".")
+ , time_buf)
+ , ".")
+ , pid_buf) ;
+
+ return (result);
+ }
+
char *SubmitFile::namefile(const char *pfix, unsigned n)
{
char buf[MAXLONGSIZE], *p;
--
Lloyd Zusman
[EMAIL PROTECTED]
