On Wed, Aug 23, 2017 at 07:40:58PM -0000, John Levine wrote:
> In article <20170731132508.GA22575@xxxxxxx> you write:
> >One of the best approaches to ensure the long-term preservation of these
> >archives is to (1) keep them in an open, standard, widely-supported
> >format -- "mbox" is clearly the best choice
>
> Er, not really. See RFC 4155. There are way too many not quite
> compatible versions of mbox. If I wanted a file format for a
> multi-file archive, I'd probably use POSIX ustar which is an actual
> standard.
Sure, but that would not be a complete spec because it does not cover
the format of individual messages nor any semantics of file names.
Whats the least restrictive sufficiently well specified generic mail encoding ?
Anything that would also pass via rfc822+mime ?
Might also be nice to have some well-known-filename with metadata
about the archive content, eg: self-URL, previous, next URL, contact
URL (if any), archive owner signed checksums of files. Template
spec of filenames (%YYYY-%MM.mail),...
> On the other hand, I see that there are rsync-able archives in some
> mbox variant, so that horse left the barn ages ago. The archives are
> kind of funky, e.g., the file dnsop/2001-01.mail contains messages
> from March 1999 through January 2001, and there are empty files
> 1999-06.mail through 2000-12.mail.
See above: metadata that would help you figure out if you have a
complete set of archive files.
Cheers
Toerless
> R's,
> John
>
--
---
tte@xxxxxxxxx