In the long term future, it may be worthwhile to move Polkadot's identity functionalities to a common-good parachain. In Polkadot, a common-good parachain has a strict definition, meaning it does not create any new tokens, and should serve for some common-good functionalities.
An identity verification, by its definition, has to interact with the real world. As a result, sometimes we need to rely on existing public-key infrastructures (PKIs). In general, those PKIs are created by trusted third-parties, especially governments, such as the Estonia's eID program. Those allows certain information to be verified (such as legal name), while not revealing any sensitive information of the verifiee.
What's more, even our "trustless" verifications rely on an implict trust to the global TCP/IP infrasturcture, and domain name services.
Right now, Polkaregistry only has one trusted PKI -- the Estonia eID's PKI. In the future, as the PKI list grow, users of Polkadot should democratically decide PKIs they want to trust and don't want to trust. This requires we put the PKI roots on-chain, in which a common-good parachain design would be really beneficial.
Polkaregistry is trustless. You don't have to trust the registrar to confirm the validity of the identity. This is already a major step forward compared with existing Polkadot registrars. In the mean time, this is still not enough. Eventually, we should move from trustless to truly decentralized.
We do this by building the common-good parachain, and let the validator set assert on the off-chain information's validity. This works similar like an on-chain oracle. The identity verification, first and foremost, remains trustless. In addition, it becomes truly decentralized, in that the seal of the identity is produced by on-chain consensus.
The role of registrars will not go. At this moment, a registrar can only attest the validity of the full set identity of an address. In the future, registrars will become more specialized and attest identity items. Some registrars may just be like now -- attesting only people who they already know. Others may serve as addition to the trustless verifications. For example, from lawyers and other professions.
At this moment, setting and verifying an identity on Polkadot is not cheap. An identity would require ~20 DOT deposit fee to set on-chain at this moment, and a verification transaction would cost a registrar ~0.015 DOT for the transaction fee. The number is likely to decrease in future Polkadot runtime upgrades. However, as the network is utilized more and more, the fee is eventually going to be similarly expensive. By moving the whole identity verification functionality to a parachain, we can off-load this burden for Polkadot, and make it affordable for everyone to set identities on-chain.