> Most orgs need to expose a lot more than that.
No, in fact there are only 2 kinds of connections:
1) public, necessarily allows anonymous/unauthenticated connections
2) non-public, doesn't need to allow unauthenticated connections
Name a service, and I can tell you which one it is.
Or put another way, there are authenticated and unauthenticated services. If you can authenticate someone at the application level, you could do it at the network level.
Most places have combined them at some boundary, for ease of access, but there is no case in which an authenticated application endpoint (even /login) has to allow unauthenticated network access to it.
And in all seriousness, name a service that is not a public website, but cannot use mTLS or IP allowlisting.
SFDC? mTLS. Customer webhooks? IP whitelisting. Data warehouse? PtP VPN.
They do it all the time, piecemeal, that's part of what anything-as-a-service delivers. If you don't want to secure an app, buy a SaaS one. Don't want to secure infrastructure?, buy an IaaS.
> Most places have combined them at some boundary, for ease of access, but there is no case in which an authenticated application endpoint (even /login) has to allow unauthenticated network access to it.
> And in all seriousness, name a service that is not a public website, but cannot use mTLS or IP allowlisting.
Literally any SaaS software where the customer does not or cannot practically install a certificate for you or provide a list of IPs in advance? mTLS is almost never used in a browser unless you are connecting to a government website with a pre-shared smartcard. IP whitelisting is great if the end users only work from behind a VPN or in an office. But there's a whole big world of people who cannot or will not do that.