[Fwd: I-D ACTION:draft-hall-mime-app-mbox-00.txt]

[Eric A. Hall]

On 5/18/2004 9:24 AM, ned.freed at mrochek.com wrote:

> The choice of CRLF as the preferred line terminator for email was made
> long ago. It cannot be changed now, and specifying types that conflict
> with this is a really bad idea.

Of course this statement is correct in multiple contexts. But we're not
talking about "mail" here -- mbox files are mailstore-specific databases,
and not collections of "Internet mail" in general. They need to be treated
in the same way that (say) Eudora or OE databases would be handleded.
Whether any specific mbox file happens to contain sequences of Internet
messages and is easy to parse as text is somewhat irrelevant, since it
could just as easily contain locally addressed messages of undeclared
eight-bit character data. The database needs to be treated as opaque, and
the end-point applications that use it should be responsible for any
necessary prior agreements and arrangements.

> Mbox files are mailed around using 7bit and 8bit encodings all the
> time. A media type that is incompatible with this very common usage is
> either (a) Not going to get used, making it worthless or (b) Will be
> used in a way that violates your MUST, which will break
> interoperability.

I can make explicit allowances for that if it would make people more
comfortable. Since the database content has to be negotiated out-of-band
anyway ("download the archive of message/rfc822 objects", or "these
messages are all big5"), folks certainly should be able to match an
appropriate encoding to the content.

-- 
Eric A. Hall                                        http://www.ehsco.com/
Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/