Re: Magic is useless!

[Maciej Stachowiak <mjs noisehavoc org>]

On 07Jan2002 12:24PM (-0500), Daniel Veillard wrote:
 
>   Yep, IMHO Mime-Type are slowing falling into obsolescence due to
> composition and the fact that it's such a f...g pain to update that
> registry that some new format tend to not even bother.
> 
>   The goal of the Mime-Type was to associate processing tool with
> resources. Unfortuantely it's a too limited view to cope with most
> of the complex formats.

Actually, it's still a very important need. Given a file, you need to
know what applications can open it. 

I'm well aware of the limitations of MIME (for example, mime types
don't let you have arbitrary levels of subtyping, which makes it hard
to support both generic XML-based applications and ones specific to a
particular kind of XML document). But it's the standard, it's what you
get from a web server, it's what you get in mail, etc.

I'm all for a better file typing system, but I don't think GNOME can
invent one by itself.

> A better approach is a list/hierarchy of strings and not a single
> one, the list of namespaces name of a XML document, the list of
> mimetypes of a zip. To take the example of a ZIP format it could be:
>
>        (zip (image/gif,   xml/docbook,   image/gif))
> 
>    to use a LISP like syntax. For a compound compressed XML document
> 
> (gzip (application/xml (http://www.w3.org/2001/svg http://www.w3.org/)))
> 
>   I'm pretty sure you can find examples even outside of the XML world.
> don't limit yourself to known-to-getting-obsolete mechanism when defining
> new interfaces.

It sounds like the pattern-matching needed to match these kinds of
types to application capabilities will be tricky to get right. In
particular, you probably want to offer the most specific app first,
but it might be hard to define what most specific means.

 - Maciej