[Top] [All Lists]

Different approach to defining encodings

1991-11-06 11:21:48
So far, this working group has been considering encodings to be
defined as a set of (in some ways) "parallel" attributes.  For
example, an object is:

   encoded-in-Base64  AND  checksummed  AND  it's-fax-in-TIFF-format

However, an object is actually a "serial" encoding which converts it
from a concept (like "noise in the air") to a mailable body part
(lines of text).  For example:

   it's-a-fax  WHICH-WE  encoded-in-TIFF-format  WHICH-WE
         added-a-checksum-to  WHICH-WE  encoded-in-Base64

Suppose we define the encoding description as a set of encoding
elements which form a statement of the algorithm(s) used in the
encoding process, and, IN THE ORDER NEEDED TO DECODE the object.

The above might be described, for example (and the syntax is only an

    base64 | crc-16 | tiff | fax


    uuencode | lzw | tar

defines an "tar" file which has been compressed and then uuencoded.

Mail user programs can now mix and match elements as appropriate to
the data and recipients.

As pointed out above, checksums can be encoding elements.  So, possibly
can characterset.

    iso-8859/3 | text

defines a text body part (where text is defined to be UTF/10646) which
has been encoded into 8859/3

Encryption can also an encoding element, although I suspect that while

    uuencode | des-56 | text

would be OK

    uuencode | des-56 | lzw | crc-32 | tiff | fax

is giving away too much information and that you probably want to

    uuencode | des-56 | message

and put the rest in the encoding in the encrypted message.


<Prev in Thread] Current Thread [Next in Thread>