Hi Terin,

On the surface the following appears to be syntactically convenient:
license-expression = 1*1(simple-expression / compound-expression / "NONE" / "NOASSERTION")

but semantically incorrect. Let me try to explain using a database data field analogy.

NONE and NOASSERTION are defined at the SPDX document field level and even have different semantic values with respect to different fields. SPDX fields can have several different types assigned which is analogous to a database field. For example, a database field may contain values of *type* Character, Integer, Float, Boolean, Date and so forth. A database field can also contain the special value NULL, which does not belong to a specific data type, but instead represents a special field value (NULL = missing unknown data). NONE and NOASSERTION are analogous to NULL, where a license expression is analogous to a type such as Character, Integer, Float, Boolean, Date and so forth.

A license expression represents the licensing terms of a piece of software (source or binary). In less precise terms, it represents the distribution obligations for a software component. NONE and NOASSERTION do not semantically represent that. To include NONE and NOASSERTION as validate license expressions is analogous to adding NULL to a database type which would be semantically awkward.

In summary, on the surface the following appear syntactically convenient:
license-expression = 1*1(simple-expression / compound-expression / "NONE" / "NOASSERTION")

But semantically it is not correct, and therefore, NONE or NOASSERTION should not be included in the ABNF definition for a license expression.

At least that is one perspective.

Best,
- Mark

Mark Gisi | Wind River | Director, IP & Open Source
Tel (510) 749-2016 | Fax (510) 749-4552