V10/vol2/upas/geoff

From arpa!geoff Wed Jul 19 00:26:22 1989
From: geoff@utstat.toronto.edu
Date: 19 Jul 1989 0026-EDT (Wednesday)
To: research!ches

Comments on the draft Upas paper of July 18.

Local authorities probably include Doug McIlroy and Dennis, so don't
take my word if you are skeptical.

Punctuation.  I tend to use hyphens to join words of a compound
adjective, as per The Elements of Style. I would suggest
"network-specific mailers" (at least thrice), "regular-expression-based
mini-language", "home-grown networks", "well-understood, and thereby
reliable mail system", "very short-lived", and "format-specific
mailer".  In "A Comparison With Sendmail", I would punctuate thusly: "We
wanted a system that had simpler, and therefore more easily verifiable,
rewriting rules." (I might also put a hyphen between "easily" and
"verifiable"). In "Message Routing", there is a period after "translate"
that should be a comma.

Spelling.  "envelope" needs a final "e".  "variant" contains no "e"
(look for "varient").

Grammar.  To get agreement in tense, I would change "there are many
implementation changes" to "there have been many implementation
changes".

Facts.  Reference 2 (uucp) claims "Unix Programmer's Manual, Seventh
Edition, Volume 2, Bell Laboratories, October 1978".  Maybe you guys
have a pre-release, but my volume 2 says "January 1979".  I recall the
release of Seventh Edition in ~June 1979 just before Usenix (I had just
phoned Irma Biren to be put on the "V7 interest list" and she said
`Guess what?  It's just been released today').

Figures.  Figure 1 shows the progression "convert -> queue -> protocol",
yet I believe the current SMTP progression is "queue -> convert ->
protocol" (isn't message-format conversion done at the last moment
now?).  Figure 2 might (again) mention message-format conversion as
happening in smtpd and smtp.  For symmetry, you might want to show the
smtpqer -> queue -> smtp progression on the left side of Figure 2, so
that one can read the flow of mail in a given protocol family down the
page.  Also in Figure 2, there is a missing or invisible line from
/usr/lib/upas/route to the arrow-head attached to smtpqer.  It isn't
clear to me how much of the diagram is covered by the name "Upas"; uucp
and relations are clearly excluded, but what about smtp and relations?
This comes up later, in comparison with sendmail.

When describing \l, it might be worth explaining how "The name of the
local machine" is determined (hostname(1), gethostname(2), uuname(1) -l,
uname(2), etc.), even if just to say "in a machine-dependent way".

After "If the command does not result in mail delivery", it might be
worth parenthetically noting that exactly "|" and ">>" perform mail
delivery.

In "Message Format Conversion", I would be interested to know how header
information is used in deriving SMTP destination addresses.  "Outgoing
SMTP messages must have at least the minimum header information required
by RFC822"; what if the don't?  Are they dropped, is the missing
information filled in automatically, or is something else done?

Re "Concealing Machine Names": I still get mail from dutoit.att.com and
arpa.att.com, so it might be worth mentioning that machine-hiding can be
fooled without trying.

Under "Installation" appears the phrase "most major versions of Unix
Time Sharing"; is that what you intended, or should it read "most major
versions of the Unix Time-Sharing System"?

The third item of "A Comparison With Sendmail" makes sense if "Upas"
excludes all auxiliaries (translate, smtp and relations, etc.);
otherwise it is clearly false: Upas now possesses most of the features
named.