From: Walter Alejandro Iglesias <wai@roquesor.com>
Subject: Re: mail(1) set Date and User-Agent [was: Re: Back to rfc2045]
To: ori@eigenstate.org
Cc: op@omarpolo.com, tech@openbsd.org
Date: Mon, 5 Aug 2024 09:20:37 +0200

On Sun, Aug 04, 2024 at 05:40:37PM -0400, ori@eigenstate.org wrote:
> Quoth Walter Alejandro Iglesias <wai@roquesor.com>:
> > Now at last appears one and takes
> > the simplest of my diffs and completely rewrite it.  What can I expect
> > will happen to the rest of what I did? :-)
> 
> That's the point of review: for someone more exprienced with the
> system to reshape the code into something they're happy to maintain.
> 

Are you the one maintaining mail(1)?  If so, and if you are interested
in these improvements, you could have done it yourself.  ;-)


> Are you interested in the featurs, or are you more attached to the
> code being untouched, the way you initially wrote it?
> 
> 

Features?  Attachment?  I'd cut my own arm if it bothers me!  My main
concern about people is honesty, about software what gets the job done.

I *test everything*, what the professional and the amateur, the
intelligent and the foolish tell me, the argument that seems solid and
the one that sounds crazy.  And above all, what my intelligence,
experience and creativity suggest to me.  My conclusions and my methods
are the result of this learning process.  In this case, taking into
account that this is not my home, and I am not an expert on the subject,
trying to adapt, I listened to many pieces of advice and ideas only
trying to make the developers of this project "happy".

Even if I were the worst C noob and half-brained considering that I have
been trying this and that for almost a year, what I've done deserves
more that *zero* credit.  Last year, I contacted you privately because
you had been the only one who had expressed interest in my proposal, but
when I had already moved on and discarded features of my original
idea, for example the use of the (-m) option, you posted a patch that
ignored my subsequent changes.  What we now see would have been a
mistake.

Regarding the date() function I found out that NetBSD doesn't have a
strftime_l() so if, as you said, portability is important, there's
little difference between Sven's solution and mine.  What I assume
should be portable is to simply do this:

size_t
date(char *str)
{
	struct tm newtime;
	time_t ltime;
	static char buf[32];
	size_t n;

	setlocale(LC_ALL, "C");

	ltime = time(&ltime);
	localtime_r(&ltime, &newtime);
	n = strftime(buf, sizeof(buf), "%a, %d %b %Y %T %z", &newtime);
	snprintf(str, sizeof(buf), "%s", buf);

	return (n);
}

At least under Linux, whatever they may use setlocale() in any other
place in the code it won't affect this function.  However, this is not
the first time I heard from a OpenBSD developer that it's better to
avoid depending on locales (I'd appreciate some explanation about it.)
If this and portability could be issues then your proposal is the option
to follow.  But, given the complexity that comes with copyright, I don't
know how far the balance tips one way or the other.  Which brings me to
the following: as I've mentioned in other threads I've changed the
function I originally wrote to generate the Message-ID to the one
OpenSMTP uses (also trying to make OpenBSD developers "happy".)  If
we're willing to keep this function, should we include it with date() in
a same file, under the same copyright?  See?  This, and what I explained
earlier about the '-m' option, suggests that doing things in chunks
directly is error-prone and a waste of time, you need to look at and
plan the whole idea first.


-- 
Walter