[Top] [All Lists]

Re: Another MIME test

1991-12-30 15:44:19
I have not had a chance to read the current specification for the new
MIME stuff.

By all means do read the document.

The concern that has popped up for me in respect to our environment will
be small messages with FAX images inserted.

My concern given my lack of knowledge of FAX protocols is that between
BASE64 and the G3 specifications, we can easily end up with length

If G3 is fairly well compressed then I have no complaint, if it is not
then I (representing my environment) see a serious problem.

Well, you have a problem here. G3 is fairly well compressed. However, the
NETFAX work is not using raw G3, it is using TIFF. TIFF is somewhat better
than raw G3 in some cases. I don't know of any cases where it would be worse.
(There is some small overheader associated with using TIFF, but this, I think
is always insignificant.)

However, if you are going to send around image data, you'd better get used to 
the idea that it is going to increase message size. While good compression
techniques can alleviate this somewhat, the bottom line is that even when you
use truly stellar compression techniques, images are big. Period.

BASE64 is a triviality. It adds a constant overhead of 33% to messages. If this 
costs you too much you can elect to set up transports for raw binary and use 
that -- MIME fully supports this. This cannot be support using present-day 
SMTP, but work is being done to fix that. But even if you do this it won't
make enough of a difference. Images are big.

Bottom line -- you're caught between a rock and a hard place. If you want
to be able to send and receive images you must be prepared to accept the
increase in bandwidth utilization this entails. If you cannot accept the
increase you cannot reasonably use images. MIME cannot help you on this. MIME 
is simply the container and labelling scheme. Don't shoot the container!

Our first application will definitely be fax messages since we have a
significant demand for that service as compared to other things like
voice, random images, etc.

If you have a problem with the specific encoding chosen by the NETFAX working
group and the compression it uses you should take this up with them. I happen
to think that TIFF is a reasonable tradeoff between efficiency and compression
costs. (Other approaches, such as JPEG, cost too much to implement casually
and should be reserved for color images and other things that need them.) I 
suspect that most of the other NETFAX folks will agree with me on this. 
However, this is not a matter for MIME or ietf-822, since we're simply blessing 
what the NETFAX folks have chosen.

Note also that MIME defines way to move around references to images rather
than the images themselves. If your environment can be made to support this
you can then use the network protocol of your choice to access the actual
image data.

If you want to propose some alternative to the schemes we've already blessed
in MIME, by all means do so (you'll have to read the document first, however).
There's always room for technical innovation in the specification. Cutting 
things out that people definitely want to use because your local network
cannot cope is not an acceptable approach though.


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