On Thu, May 15, 2025 at 02:25:44PM -0700, Will Yardley wrote:
> https://www.loc.gov/preservation/digital/formats/fdd/fdd000386.shtml
> mentions MBOXCL and MBOXCL2 dating back to SYSVr4. So, if that's the
> case, that may have been the widely understood "mbox" used well before
> Mutt was first written?
Informally, "mbox" format dates back to the fifth edition of Bell Labs
"Research Unix" from 1974. The oldest mbox files that I still have are
from 1980, and while I haven't tested them recently, they *were* readable
with mutt a few years ago. We referred to them as "mboxes" back in the
day; the term was so common that some programs used "mbox" as a default
filename. (Can't recall offhand, but maybe the early BSD version of
the "mail" command, AKA "ucbmail"?)
Formally, "mbox" format is much more recent, thanks to RFC 4155:
RFC 4155 - The application/mbox Media Type
https://datatracker.ietf.org/doc/html/rfc4155
which dates from 2005. But here's the thing: that RFC was written
(pretty much) by looking at extant mboxes and working backwards, which
is why it begins:
"The mbox database format is not documented in an authoritative
specification, but instead exists as a well-known output format
that is anecdotally documented, or which is only authoritatively
documented for a specific platform or tool."
But since some standard is (usually) better than no standard, and it looks
to me like RFC 4155 was well-researched, it's probably worth using as a
reference.
I was unaware of the all the variants in the mutt ecosystem until this
discussion, so I don't know which comply with RFC 4155 and which don't.
(Hopefully they all do. It would probably be a good idea to check.)
But I'll toss this into the discussion for anyone who's working on them:
it's really, really important not to break backwards compatibility.
Because despite whatever faults it might have, mbox format is *the* de
facto open standard for mail archiving, ahd there are a lot of things
that depend on it -- including some very long-term projects, and including
tools like "grepmail".
---rsk
