There's no doubt that the BBS Signature was born from this classic Short Group Signature paper in 2004. It's capable of Zero-knowledge. In the paper, section 5, it describes how a member of a group, can sign a message, without revealing which member signed it. But, this group signature can be verified that it was signed by a member of a group (anonymously) through a group public key $gpk$.
However, later, there are some works improving BBS into BBS+. It was first modified and named in Constant-Size Dynamic k-TAA by Au et al. BBS+ was touched again in Anonymous Attestation Using the Strong Diffie Hellman Assumption Revisited. Nevertheless, in those two papers mentioned above, it seems the BBS+ scheme is a multi-messages signing protocol. It takes messages $m_1, m_2, m_3,..., m_L$ and produces a single signature $\sigma$. It can create a proof of knowledge $SPK$ on top of signature $\sigma$, with some messages $m_i$ revealed.
Yet, in work on Enhanced Privacy ID from Bilinear Pairing by intel, it seems their BBS+ scheme becomes as a member of a group signing the message anonymously, same as Short Group Signature paper in 2004.
Therefore, I'm confused about BBS+ protocol. Whether it is
- a multi-message $m_1, m_2, m_3,..., m_L$ signing protocol producing a single signature $\sigma$ with signature proof of knowledge $SPK$
- a group signature signed by a member of a group anonymously using a group public key $gpk$?