My understanding is that CRL is a list from a CA that says what certificates not to trust.
Browsers will regularly download these lists from their trusted CAs and check the certificate of requested sites against these lists, alerting the user if the certificate has been revoked.
It seems to me that the process of downloading the (probably) millions of revoked certificates from CAs and searching that list every time the user connects to a site would be quite slow. Am I missing something in this protocol?