[standards-jig] sending mime-type during filetransfer

Tomasz Sterna tomek at smoczy.net
Mon Jan 12 17:08:42 UTC 2004


Richard Dobson wrote:

>>>Yea it is but I still dont really see the requirement for mime types in
>>>filetransfer because their is not the requirement to be able to
>>>      
>>>
I am trying to forecast a problem here...

A binary chunk of data is preety useless without information of the type 
of the data.

Todays OS'es relies on the so called "extension" of the file to guess 
mime-type of the file
or the magic-file to determine the type from the contents of the file.
It's a very error prone method.

But modern OS'es like BeOS, AtheOS, Cosmoe, Syllable, MacOS etc. stores 
the mime-type
alongside the file in so called "attributes" and use them to determine 
the file type. Name of the file
is purely human oriented. There is no such concept of file "extension" 
there.
The web servers are doing the same thing. And it's the way to go.

Sending mime-type alongside the binary stream is usefull. The sender 
almost certainly knows what is
the type of the data it is sending. And guessing it on the reciever side 
can be misleading.

If not making it a REQUIREMENT like in HTTP, we could at least make it 
an OPTION.
If  the reciever "knows better", it can ignore the type, and try to 
guess it using an extension,
magic-file, crystal ball, anything.


P.S. Yes, today I'm in the mood of getting to my mailbox using Mozilla 
Thunderbird running on Windows XP.
Does that make what I said untrue? I was using Evolution running on 
Linux the other day, or mutt running
on AtheOS. Does my mail headers really are in the topic?




More information about the Standards mailing list