From mboxrd@z Thu Jan  1 00:00:00 1970
From: steffen@sdaoden.eu (Steffen Nurpmeso)
Date: Mon, 12 Sep 2016 14:34:56 +0200
Subject: [TUHS] BSD mail(1) line addressing
In-Reply-To: <20160912004332.4362323.94475.6921@stanleylieber.com>
References: <20160912004332.4362323.94475.6921@stanleylieber.com>
Message-ID: <20160912123456.6xQx8sec1%steffen@sdaoden.eu>

stanley lieber <sl at stanleylieber.com> wrote:
 |Anyone know what considerations informed the decision (way back when) \
 |to eschew ed-like line addressing and switch to the currently entrenched, \
 |reversed 
 |form?
 |
 |EXAMPLES
 |
 |V8 mail:
 |
 |1,5p # prints messages one through five
 |
 |BSD mail:
 |
 |p1-5 # prints messages one through five

I cannot see any code that would parse the command line, let aside
an ed-style one, in V7 usr/src/cmd/mail.c.  But interesting
question, why that prefix-based command syntax has been thrown
over board.  Maybe, even likely i would say, Kurt Shoens didn't
even do so, what the question implies, but instead the message
specification is the actual argument for the command, instead of
being just an address constrained of it.

For some commands this is a restriction, however, which is why the
laststring() (former getcmd()) has been introduced in the nail
fork.  It backward parses something off the end of the command
line, to overcome ambiguities that would otherwise arise, e.g.,
the `pipe' command takes a message-list and a command.  This is of
course a very bad hack.  Parsing the message-list should be
prepared for not failing once the shell command is reached, and
the shell long has the "--" mechanism for such separating
purposes.  nail(1) has not, maybe mailx(1) will, one day.

--steffen