Incorrect MS Office mime type detection

[Mike Hearn <mh codeweavers com>]

Hi,

We have a slight problem, in that the shared mime info database appears
to detect all Office files as MS Word Files. This leads to the
all-too-familiar security warning and inability to open PowerPoint or
Excel files associated via Wine with Office from the file manager.

I'm getting concerned at the number of hacks being put into gnome-vfs to
special case unusual MIME detection scenarios. I've been proposing a
pluggable sniffer system for a while now but never sat down and specced
one out, so these sort of problems are just going to keep cropping up
with formats that cannot be easily differentiated by simple binary
pattern matching.

For now, I'd appreciate it if the next version of shared-mime-info gave
precendence to the file extension in the case of OLE Structured Storage
streams, as if there is a way to tell Word vs PowerPoint vs Excel files
apart by magic sniffing I don't know what it is. 

I think it's not too hard to write a program that parses the stream
format to figure it out but there are no hooks we can drop such a
program into.

We'd really appreciate if this problem can be solved ASAP, ideally in
time for Fedora Core 1 or the next round of updates that can be pushed
to it. Otherwise our users will find it really awkward to open Office
files from Nautilus due to the wrong security checks.

thanks -mike