More streamlining
2006-02-23 19:21I've been discussing my ideas on the Lemonade mailing list.
Unsurprisingly, the gurus there didn't like my attempt to rehabilitate smtps, mainly for political reasons. The RCPTHDR extension probably only has a future as input to the RFC 2822 or RFC 2476 updates.
People seemed fairly keen on QUICKSTART. I have updated it to shave off a couple of round trips from connections that use STARTTLS.
Further analysis of the BURL draft revealed that its pipelining changes aren't totally correct - see my message on the subject. The upshot is that you can't save a round trip for AUTH.
However, a closer look at the CHUNKING spec revealed that it has a very desirable property: it completely eliminates synchronization points once you have got the connection going and you are sending messages. With basic ESMTP, there is a synchronization point at the DATA command, which has the effect of limiting your sending speed to at most one message per round-trip. CHUNKING allows you to send at the maximum speed TCP can sustain. Nice. The CHUNKING RFC needs a little clarification about its interaction with PIPELINING, but its author has a draft update in the works which will do the trick.
I have more to say about sending email efficiently in future essays on how not to design MTAs.
Unsurprisingly, the gurus there didn't like my attempt to rehabilitate smtps, mainly for political reasons. The RCPTHDR extension probably only has a future as input to the RFC 2822 or RFC 2476 updates.
People seemed fairly keen on QUICKSTART. I have updated it to shave off a couple of round trips from connections that use STARTTLS.
Further analysis of the BURL draft revealed that its pipelining changes aren't totally correct - see my message on the subject. The upshot is that you can't save a round trip for AUTH.
However, a closer look at the CHUNKING spec revealed that it has a very desirable property: it completely eliminates synchronization points once you have got the connection going and you are sending messages. With basic ESMTP, there is a synchronization point at the DATA command, which has the effect of limiting your sending speed to at most one message per round-trip. CHUNKING allows you to send at the maximum speed TCP can sustain. Nice. The CHUNKING RFC needs a little clarification about its interaction with PIPELINING, but its author has a draft update in the works which will do the trick.
I have more to say about sending email efficiently in future essays on how not to design MTAs.
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}