The issue of mime parsing

Hi,

I'm new to the list, but have followed it for some time and also
tried to read relevant info I could find through google.

Regarding the recent discussions in connection to the BikINI project
about server-side MIME parsing; my question is: is it perhaps
nessecary to abandon MIME for im2000? This seems to be the only
solution to both:
  i) allowing bandwith-saving fetches of only relevant parts of a
  mail (which is yet more important if I'm now fetching my mail
  from someone else's server on the other side of the globe)
  ii) avoid the pain of too intense server-side parsing.

As an alternative solution I propose that a mail is handled as a
"file system directory" which contains all the parts of the message.
The directory may then also contain an *easily parsed*
'manifest'-file with information about the role and meta-data of the
contained part-files and their file formats. (eg., this file may say
things like: 'message.txt' and 'message.rtx' are redundant 'text' and
'rich text'-versions of the same information.). Fetching only a part
of a mail is now as easy as just downloading a specific file.

In case people strongly dislike the idea of abandoning "one mail=one
bitstream", it is possible to store the mails tgz:ed or zip:ed. This,
however, makes parsing more complicated as the mail has to be
temporarily unzip:ed first.

Best regards,
Rickard Armiento