[coderobe]

>SSL is terminated at an edge location that is closest to the users.

So this is routing plain text http for most of the connection and at the same time giving the managed edge location direct access to the traffic and a valid tls cert for the domain? Isn't that mostly snake oil then, as the secure connection never originates from the target service?

[Sandeepg33k]

Hi.. The TLS is terminated at the edge, and from that point we fetch the data from origin server. As long as the origin has SSL, the communication is secure end-to-end.

[coderobe]

>As long as the origin has SSL, the communication is secure end-to-end.

It cannot be secure end-to-end, as your edge location is quite literally performing a MITM. That aside:

How are you validating the TLS cert that the origin presents?

Going by the info on your website, the possibilities are as follows:

Scenario 1: The SAAS provider presents a TLS cert not valid for customer-domain.com when accessed as customer-domain.com

Scenario 2: The SAAS provider presents a TLS cert valid for customer.saasprovider.com when accessed as customer.saasprovider.com

Assuming scenario 1, you would need to validate the certificate out-of-band as the traditional trust chain does not validate for the given domain. Assuming scenario 2, you would need to rewrite the URLs from customer.saasprovider.com to customer-domain.com to prevent the users from following generated resource URLs to the origin domain. Or am i missing something?

[bauruine]

>As long as the origin has SSL, the communication is secure end-to-end.

No it isn't. You are doing a MITM.

[philsnow]

you're talking about "piecewise end-to-end".

the "end"s are the _browser_ and the _origin_, and if there isn't a single secure channel that goes all the way between them, that's not "end to end".

I mean take the "piecewise" argument to its natural conclusion.

If the reason it's okay for you to be in the middle is that you're going to ensure that your request to origin is also encrypted, why should you be the only party in between that can decrypt the contents of the connection?

Why not let the ISP also decrypt the contents? What about the layer 3 interconnect providers? How about your cable modem and your router (they're _probably_ patched 'enough' that it's safe to let them see your plaintext).

I'm harping because misuse of the term "end to end" is _actually dangerous_ to real people.

All of this is to say nothing of the fact that when you allow "middle-boxes", the client no longer has control over the ciphers that are used for the end-to-end connection, so they lose control over perfect forward secrecy.

[philsnow]

you might say,

> but this is what cloudflare does!

yes, and it already caused one of the worst breaches in the short history of the internet https://news.ycombinator.com/item?id=13718752