Re: Suggestion for file type detection approach

[Olaf Frączyk <olaf cbk poznan pl>]

On Wed, 2003-12-31 at 00:07, Geoffrey wrote:
> Olaf Fra;czyk wrote:
> 
> > File sniffing does not guarantee it either. Users expect, that the
> > action (which application is to be run) is determined by file
> > extension.
> 
> If you're born in a windows world, yes.  I've never expected a file 
> extension to define a file type.  I refer again to your typical system 
> directories: /bin /usr/bin /sbin ...
This is another story (at least partially). We have executable
permission "x" to distinguish if we have an application or data.
And... most desktop users are Windows users.
> 
> > The mime-type idea is a terrible thing for a normal user.
> > They simply don't understand it.
> 
> Stupid users does not make a solution incorrect.  That's based on your 
> definition of a normal user, which apparently excludes most knowledgable 
> users.
If 99% users are stupid, then normal user is a stupid user. And this is
the reality.
And if a solution is not understood by 99% people who need to use it,
then the solution is incorrect.
> 
> >>> I am all for nautilus using the extension when available and
> >>> sniffing when appropriate (a file missing a suffix or when
> >>> requested by the user)
> >> 
> >> File sniffing is at least attempting to apply some logical effort
> >> to file determination.
> > 
> > But impractical because of slowness. And not ease to understand for 
> > users.
> 
> Based on your definition of 'users.'  I'd prefer to give up the time for 
> an accurate file identification.  Change the name of a file and you've 
> got the wrong identification.
> 
> > 
> > The ideal would be ti switch to recognizing file type by it's
> > extension but if one extension is used for different mime types, then
> >  in such case nautilus should detect mime type from file content.
> > This should be controlled by user - eg. if user wants to associate an
> >  extension with another application, he would get dialog box with 
> > question if previous association is to be dropped or if the mime type
> >  for this extension should be detected by file content.
> 
> I don't believe one solution or the other will please a majority of 
> users.  Relying on file extension has gotten windows into a lot of 
> trouble. I prefer both solutions be offered and then let the user 
> choose, speed and insecurity or accuracy and security..
Here I agree with you. (about the ability to choose). 
But I don't see any security problem here. The only thing that I could
suspect is hiding of extension in Windows. But it doesn't mean that
using extensions is dangerous. Only hiding them is dangerous.

Regards,

Olaf
> 
> -- 
> Until later, Geoffrey	esoteric 3times25 net
> 
> Building secure systems inspite of Microsoft
> 
>