Re: betteremail.com
<dsr <at> tao.merseine.nu>
2003-01-02 01:17:04 GMT
On Wed, Jan 01, 2003 at 07:23:03PM -0500, Dan Melomedman wrote:
>
> Also see this: http://www.emsd.org/
This is an interesting one. I've read the white paper, but not anything
else, so these criticisms are based on that and are probably incomplete
and may even be erroneous:
- EMSD uses a new UDP-based protocol to support reliable transmission of
packets. The benefits of this over TCP are not clear to me, especially
in light of the fact that TCP is well-understood and widespread by
many applications; using ESRO exposes a mail system to bugs in a new
underlying layer, which strikes me as suboptimal. The author claims
that a reduction from 9 to 3 transmissions makes this an acceptable
tradeoff. (SMTP with pipelining vs EMSD) However, if you compare QMTP
vs EMSD, the number is now 5 vs 3. That doesn't sound worthwhile to
me. Also, UDP-based protocols are generally a hassle for firewalls.
- EMSD uses RPC. I hate debugging RPC based protocols, I hate trying to
verify them, and I hate pushing them through firewalls.
- EMSD uses ASN.1. The vulnerabilities of most ASN.1 parsers are
well-known; I am willing to bet USD5 that another multi-parser flaw is
found in the next year. ASN.1 goes against my feelings about protocols
on the Net: they should be human-readable, human-writable,
human-debuggable.
Basically, if I needed a short-message delivery protocol with
guarantees, I would look at QMTP, BEEP, RSS or a custom TCP-based
protocol; if I didn't need guarantees of any sort, I would use a custom
(Continue reading)