[standards-jig] Re: sending mime-type during filetransfer

Sami Haahtinen ressu at ressukka.net
Tue Jan 13 05:37:53 UTC 2004

I'm not usually keen on taking part of flamewars, but now i just have to 
correct something here.

Richard Dobson wrote:
> Im not sure how this is going to help with the fact that the majority of
> operating systems have no place to store and use this mime type/content type
> that it is being proposed is sent along with the file. This mime
> type/content type is only useful if it will follow the file and later when
> the user tries to open it outside of the IM app use that mime type to
> determine what application to open the file in, AFAIK windows and linux will
> just use the file extension.

Ok, first of all, please describe what you consider to be majority of 
operating systems, by number or by percentage of all operating systems. 
It is clear that at this point the leading desktop operating system is 
Windows which adds up to 3 operating systems (9x/ME, NT series, 2000/XP) 
and some 6 versions, so considering MacOS, AtheOS (i heard that 
somewhere) and most likely MaxOS X supports storing mime-types, we are 
already in a 50/50 situation. I wouldn't be so sure that majority of 
operating systems ignore mime-type.

And yet another thing that needs to be corrected in your post, most 
linux applications do not rely on the extension, the extension is there 
just for convinience, it's something people have learned to do. Most 
civilized Filemanagers probe the mime-type automatically (which adds 
overhead) based on contents of the file.

And this brings me to my point, if the sending end knows the mime-type 
of the file, why should it care if the other end is stupid enough to 
ignore it. We cannot assume the extensions are the same on all platforms 
or that there will never be an extension collition. We can only try and 
help the receiving end to get as much information of the data being sent 
as possible.

Oh, and  why give it a filename at all. Think about if the file will 
never be stored on a filesystem that supports filenames (yes, there are 
filesystems like that too) the file could be given an ID which never 
gives a clue of the original filename.

Ofcourse, i'm way off the point here, but then again, so are most other 
posts in this thread, so why should i be any different ;)

-- Sami

More information about the Standards mailing list