[media-types] Primary file extension and other additional info rmation

Hi everyone,

I researched the media type topic and the different projects and use
cases. As result I started the media-types project:
https://github.com/media-types/media-types

It pulls the data from the IANA Media Types Registry and parses it to
create an enhanced CSV file. This CSV is used to generate
/etc/mime.types. I like this project to just provide what the IANA Media
Types Registry declares. I noticed that I like to have more information
in the registry:

* Human Name: A name that can be presented to the user.

Examples:
- "Word document" or "Microsoft Word document" for application/msword.
- "MPEG-4 video" for vidio/mp4.

See "comment" section in
https://gitlab.freedesktop.org/xdg/shared-mime-info/-/blob/master/data%2Ffreedesktop.org.xml.in
for more examples.

* Primary file extension(s): Only one media type can claim a primary
file extension.

Example:
- application/xml would declare .xml
- application/watcherinfo+xml would declare .wif but not .xml
- text/html would declare .html and .htm

* Glob pattern: Similar to file extensions, but expressed as glob
pattern.

- application/xml would declare "*.xml"
- text/vnd.debian.copyright would declare "copyright" (the file is named
exactly that way)
- application/vnd.g3pix.g3fc would declare "*.g3fc" and "*.g3fc[0-9]"

* Primary Glob pattern: Similar to primary file extensions, but
expressed as glob pattern.

Is this mailing list the right place to pitch this idea? What do you
think?

On Sat, 2026-07-25 at 15:10 +0200, Benjamin Drung wrote:
> Hi everyone,
> 
> Linux distributions come with a package providing /etc/mime.types that
> contains mimetypes and file name extensions. Many of those use
> https://www.iana.org/form/media-types as input for it. To avoid
> duplicating work and having small differences in the distributions, I
> like to ask IANA to extent the data provided on
> https://www.iana.org/form/media-types with two more columns: one for
> file name extension and one for magic numbers.
> 
> RFC 6838 §4.12 mentions magic numbers and file name extensions. So this
> data is already collected when new mime types are requested.
> 
> What do you think? Are you open for that idea?
> 
> I propose following specification:
> 
> A mime type can have zero to multiple magic numbers and zero to multiple
> file name extensions.
> 
> The first file name extensions mentioned is the preferred file name
> extensions.
> 
> Similar to mime types, magic numbers and file name extensions must only
> be assigned once. In case of conflicts, they should be resolved by a
> process that needs to be defined.

-- 
Benjamin Drung
Debian & Ubuntu Developer

_______________________________________________
media-types mailing list -- [email protected]
To unsubscribe send an email to [email protected]