The Debian Package Tags (Debtags) can be downloaded here:

http://debtags.alioth.debian.org/vocabulary/

In my opinion, many of these facets can be put into the Trove categorisation properties:

• Accessibility Support (Trove: Topic)
• System Administration (Trove: Topic)
• Biology (Trove: Topic)
• Culture (Trove: Topic)
• Software Development (Trove: Topic)
• Field (Trove: Topic)
• Games and Amusement (Trove: Topic)
• Hardware Enablement (Trove: Topic)
• Made Of
• User Interface (Trove: User Interface Type)
• Implemented in  (Trove: Programming Language)
• Junior Applications (Trove: Topic)
• Electronic Mail (Trove: Topic)
• Office and business (Trove: Topic)
• Works with
• Supports Format
• Scope
• Role
• Security (Trove: Topic)
• Sound and Music (Trove: Topic)
• Service tags
• Application Suite
• Network Protocol (Trove: Topic)
• Interface Toolkit (Trove: Topic)
• Purpose
• World Wide Web (Trove: Topic)
• Networking (Trove: Topic)
• X Window System (Trove: Topic)
• Science (Trove: Topic)
• Writing script (Trove: Topic)

In my opinion, ADMS.F/OSS should not mandate one particular controlled vocabulary for classifying software. It may only suggest or recommend useful controlled vocablaries to enhance interoperability. Consider the following levels of software metadata interoperability: 

1. Level 1: Every software forge or catalogue uses its own classification system, which is not documented in a machine-readable format.
2. Level 2: Every software forge or catalogue uses its own classification system, which is documented on the Web, in a machine-readable format such as SKOS.
3. Level 3: Every software forge or catalogue uses its own classification system, which is documented on the Web, in a machine-readable format such as SKOS. There exist automated mappings between these classification systems.
4. Level 4: Every software forge or catalogue uses the same universal classification system, , which is documented on the Web, in a machine-readable format such as SKOS.

​Clearly, level 4 is unrealistic, there may always be good reasons for a software forge or catalogue to use its own classification system. It is much better to spend time and energy in attaining level 2 and 3.

I agree with Stijn that it should just recommend, at least for a start.

I need arises then we can think of unification.

Hopefully, RDF is extensible, and allows mixing many different vocabularies.

Proposal: The following Debtag software classification categories seem relevant to include to the specification, as they support the main use case of searching for software.

• Supports Format  - "Which data formats are supported by the package"

The controlled vocabulary should not necessarily be limited to the once propsed by Debian. See also related proposal: https://joinup.ec.europa.eu/discussion/it-allowed-create-new-code-values#comment-12273.