smtp-recap.md

SMTP protocol recap

An SMTP conversation would typically look like this:

1. Client initiates a TCP connection to the server.
   • The MTA should talk first (client should wait!) and print a line with some basic indicators, e.g.:

     220 myhost.g3rt.nl ESMTP Postfix (Debian/GNU)
     

     where 2xx code indicates 'OK'.
   • The MTA may already reject the client at this point (e.g. IP blocklisted in RBLs).
   • Note that Postfix servers sending mail out are also SMTP clients.
2. Negotiate capabilities, e.g. upgrade connection security with StartTLS.
3. say HELO mysmtpserver.g3rt.nl (or EHLO instead of HELO)
   • The MTA may reject the connection already at this point (e.g. invalid HELO name).
4. say MAIL FROM:<userx@y.foo.org> to indicate the envelope sender (Return-Path).
   • The MTA may reject the sender address at this point (e.g. prohibit impersonation, sender address non-existence).
5. say RCPT TO:<userc@d.bar.org> for every recipient.
   • The MTA may reject the recipients (e.g. to deny relay access).
6. say DATA to proceed to sending headers.
7. say HeaderName: header value for every header.
8. say empty newline to proceed to sending the body.
9. Send the body as given a negotiated transfer encoding and policy, typically MIME-encoded with lines wrapped up to 72 characters.
10. say . on a line of its own to indicate the end of the message.

At any stage the MTA can reply to the client with a reply code, indicating extra information, a successful completion or an error.