correct • elegant • free

△ comp.mail.* △

◅ Message Delimiters

Addressing question ▻

Message Delimiters

In article <>,
Jamie Zawinski  <> wrote:
>Once again, there are existing mail files which:
> 1: faithfully reproduce the contents of the messages in them, and
> 2: happen to have lines that begin with ">From ".
>Your new format will not faithfully present those messages to the user.

This is true.  However, you have no way of identifying these mail files,
because the "mboxo" format loses information.

>Therefore, it is not a compatible solution.

By "compatible", I mean that mboxo readers understand mboxrd files,
and that mboxrd understand mboxo files.  You can even have a file with
messages written in the 2 different formats, and both readers will
continue to understand it.  You can upgrade any given writer or reader
independently of all the others.

Sure, mail messages will continue to be displayed corrupted till *all*
software has been upgraded.  I don't see this as a problem, because
messages were displayed corrupted *anyway*.

Furthermore, I claim that the sequence `From' is considerably more
common (in original messages) than `>From', `>>From', etc., so an mboxrd
reader sees fewer corrupted messages when reading an mboxo format file
than an mboxo reader would.

>You're proposing a new format, and you're saying, "no, it's not 100%
>compatible."  But your new format will cause confusion.

I don't believe that occasionally seeing `From' where you should have
seen `>From' is any more confusing than the other way round.  Certainly,
none of my users have complained since I started introducing mboxrd
format over two years ago.

>                                                         Because this
>new format has the same problem as all the other formats, which is that
>you cannot examine the file and know which format it is.

No no no.  The problem with mboxo is that it *destroys* *information*.
By using an mboxrd writer, you avoid destroying that information, and
you can subsequently retrieve it with an mboxrd reader.

Granted, there's no way to tell an mboxo file from an mboxrd file.  But
why do you want to?  If your goal is to present corrupted messages on as
few occasions as possible, then, given an mbox file of unknown origin,
your best strategy as a reader is to assume that it is in mboxrd format.
(Per my claim above, this will usually produce better results even if it
isn't.)  From this, it follows that your best strategy as a writer is
always to use mboxrd format.

[ I'm using Dan's terminology: "mboxo" is the original mbox format,
using the substitution s/^From />From /; "mboxrd" is Rahul Dhesi's
format which uses s/^(>*)From /\1>From /.  Incidentally, I invented this
independently, and first implemented it on 1995-04-04; I'm delighted
that others have had the same idea and the meme is spreading. ]

Tim Goodwin   | "USENET, of course, is a pure and unadultered source
Cambridge, UK | of truth and wisdom." -- Richard Kettlewell

Original headers:

From: (Tim Goodwin)
Newsgroups: comp.mail.headers,comp.mail.misc
Subject: Re: Message Delimiters
Date: 9 Aug 1996 13:52:17 GMT
Organization: Unipalm PIPEX
Message-ID: <4uffqh$>
References: <4r16vn$> <>
  <> <>

△ comp.mail.* △

◅ Message Delimiters

Addressing question ▻