Consider the consequences of compromise of a secret key in the Kerberos system vs. in SSL. For example, suppose your individual shared secret key (for your user account) becomes compromised (i.e. it is learned by someone else) in Kerberos. And suppose a widely used certification authority's (CA:s) RSA private key becomes compromised. In which system is it easier to revoke the compromised key (i.e. prevent its being used by an adversary who obtains it), and why?
Kerberos is an on-line mechanism that is centralised. You could in principle change your Kerberos password instantly. (There might be further problems if the attacker beat you to it, but let's assume you could phone up support to change it in this case.)
Tickets issued by Kerberos have a limited lifetime (e.g. 12 hours, although this is configurable), and also have a limit in the number of times they can be renewed (which can be configured in the KDC). Here are typical settings in Active Directory. A compromised secret would have a limited effect, within the lifetime of existing issued tickets.
In contrast, PKI are intended for more distributed usage, where clients and services don't have to contact the CA every time one of its certificate it used; checking for certificate revocation in a PKI (via CRL or OCSP) tends to be an optional step. Not all clients do it. Certificate also tend to have a longer lifetime (typically one or two years), which would mean that a client failing to check for revocation would still consider a compromised certificate as valid for a much longer period of time.