I'm the assigned ARTART reviewer for this draft. I have little prior familiarity with RPKI and ROAs but have a few broader comments.

1. idnits found several issues with the draft. 16 lines are too long, which caused them to be cut off on the rendering I read, most importantly in the ROA formal definition in Section 4. [RFC6268] is referenced at line 180, and needs to be included as a reference.

2. Section 4.3.2 says that ROAIPAddress contains IPAddress elements while (if I'm reading the definition correctly) it contains BIT STRING elements. RFC 6482 defined an IPAddress element as a BIT STRING but that definition was removed; perhaps the text didn't keep up with that.

3. RFC 4291 may be a normative reference due to the MUST NOT in Section 4.3.1.

4. I couldn't understand how to do the comparison in Sections 4.3.3 and 4.3.3.1. Specifically, what is the precedence between afi/addr/plen/mlen? It seems like that might be the order, but I don't see it specified anywhere (other than perhaps in the example implementations, which are informative).

Otherwise, the document is clear and well-written.