I took a quick glance at the draft from IoT point of view (only partial review ). 

 

Section 4.2 talks about the IoT usecase as Direct Client Mode -- where they talk about cheap devices which may not have unique UUID associated with it. 

Note that a client that operates as above that does not have a
   globally unique link-layer address on any of its interfaces MUST NOT
   use a link-layer based DUID (DHCP Unique Identifier), i.e., DUID-LLT
   or DUID-LL.  For more details, refer to Section 11 of [RFC8415].
 
1. However,  it is not clear what  source initial link-layer address should be used by these devices. should it point to section 6? will that suffice?

2. Moreover,  how safe the mechanism would be if the Security section says that mechanism defined in this draft may be used by a bad actor ? 

3. It appears to me the mechanisms are designed for VMs behind an hypervisor and then IoT usages are added. My concerns are two fold for challenged low capability IoT devices -- 1) will they be able to handle the complicated option processing described here? 2) How to mitigate the security vulnerability for IoT devices as direct clients?  (The security section does not talk about mitigation)

Should there be a simpler option processing structure without TLV option processing ( i,e a fixed structure part + then TLV part for optional information]?