FreeLists / openbeos / [openbeos] Re: kernel mime_table module

["David McPaul" <dmcpaul@xxxxxxxxxxxxxx>]

On Sun, 09 Jan 2005 18:13:26 +0100 Axel DÃrfler <axeld@xxxxxxxxxxxxxxxx
> said
> Ingo Weinhold <bonefish@xxxxxxxxxxxxxxx> wrote:
> > I wouldn't object to something like a 
> > BNodeInfo::GuessMimeType(BMimeType 
> > *), which would take the file name into account, if the data-based 
> > sniffing 
> > is inconclusive (just like the entry_ref version of 
> > BMimeType::GuessMimeType()).
> 
> Shouldn't it go the other way around? Ie. test the name first, and 
> then 
> look for the contents?

The trouble with testing the name (usually just the extention) is that 
you run the risk of files being disguised.  Extensions are all well and 
good as long as they can be trusted.  As any windows user can tell you, 
extensions cannot be trusted.

I would ask the question, Can we guarantee that an executable file 
ending in .AVI will never get executed by the OS.

This is particularly important if we guess first from the extension 
then get it right from the contents which may involve a delay where the 
user can be fooled into running it.

User sees AVI extension and ICON says its a movie.  Goes to double 
click, OS finishes sniffing file and changes icon to executable.......

> BTW BNodeInfo::GuessMimeType(BMimeType *) should have a second 
> parameter like "filenameOnly = false" or "sniffData = true" - that 
> way 
> it could directly be used by things like Tracker to get a rough guess 
> of the file type as fast as possible (it then would just show the 
> generic file type icon for those where this fails, and could do a 
> deeper test for the data asynchronously).
> The advantage of this approach is that all file types where the 
> suffix 
> is sufficient are determined immediately with no visual delay for the 
> user. Since this is only important for non-BFS volumes anyway, the 
> chance that the suffix is sufficient is a very good one, too.
> 
> For mimetype enabled volumes, BNodeInfo::GuessMimeType() should 
> probably store the type, too - it doesn't make a lot of sense to lose 
> it again then.
> 
> Anyway, the most important thing to get right is BRoster::Launch() 
> anyway, since it's quite annoying for the user (while the missing 
> icon 
> is just unfortunate).

Cheers
David


--
Brain, n.:
        The apparatus with which we think that we think.
                -- Ambrose Bierce, "The Devil's Dictionary"