DOCBOOK: Re: DocBook filename extension

[Steinar Bang <sb at dod dot no>]

>>>>> Brian Lalonde <brianiacus@yahoo.com>:

> From: "Steinar Bang" <sb@dod.no>

>>>>>>> Brian Lalonde <brianiacus@yahoo.com>:

>>> When is a MIME type more useful than an extension?

>> As to "When", my answer is "always".

> Not so!

> Proof by existance counterexamples:
>   1. text/plain  .txt .c .h .pl .pod .java .doc etc

This is because no-one has bothered to register the source formats .c,
.h, .pl, and .java MIME types yet.  They probably should be text/*
subtypes.  Maybe there hasn't been a use for them yet?  I suspect
there will be a use for them when/if file systems show up, that have
the IANA meduia types type system built in.

(A counter-counter-example: is a .h file a C or a C++ file?)

POD should probably have been registered as text/pod.

I hope you don't send .doc as text/plain.  That would make things
awfully messy in both MUAs and web browsers.

>   2. application/octect-stream  .exe .class .bin

application/octet-stream, means basically: transfer "this file without
changing it in any way" (reversible transfer encodings are ok).

>   3. XSL: text/xml or text/xsl?

I think XSL should have a MIME type of its own.

Others think otherwise, because they think all XML can be handled by
parsing into a DOM and applying the appropriate magic.

>   4. text/xml files (XSL, RSS, et al may not include a !DOCTYPE)

I personally think that all well defined XML formats should have a
media type of their own.  Others disagree.  See 
	<http://www.faqs.org/rfcs/rfc3023.html>
for the definition of some XML and XML releated formats.

>   5. The degenerate application/* (which is not always centralized)

Proper media types are always centralized and can be found here:
	<http://www.isi.edu/in-notes/iana/assignments/media-types/media-types>

Experimental media types are prefixed with "x-".  Illegal media types
are unregistered media types that aren't prefixed with "x-".

> Aside from .doc, I cannot think of any examples of MIME types that
> provide greater specificity than extensions.

That's probably because you weren't trying too hard. :-)

.idl is one (there are more than one "Interface Definition Language")
.mdl for:
 - Digitrakker Music Module
 - Rational Rose models
 - Quake model files
 - 3D design plus module
 - Simulink Simulation Model
 - CA-Compete! Spreadsheet

Please see <http://filext.com/> for many examples.

> Neither system is perfect, but extensions, in my experience anyway,
> have been less vague.

Only IANA media types can IMO be called a system.  The other is a lot
of unconnected ad-hoc bindings of file name extensions to files used
by different applications.

[snip!]
> Again, aside from .doc, I cannot think of other significant examples
> of extension collision.  MIME's structured, categorized approach is
> attractive,

Yes.

> but there seem to be too many generalizations (for me, anyway),

Like eg. what?

> and inconsistencies

Hm... what inconsistencies?

> (I can never remember which MIME types use ".vnd", or "x-").

I thought .vnd was for undocumented formats, but the registry has dem
for a latex format, so then I guess I don't understand what it's for. 

x- is for "experimental".