On 30/07/2014 15:07, BlueStar88 wrote:
> Am 30.07.2014 um 14:17 schrieb Daniele Nicolodi:
>> One of the main features of the current email infrastructure is its
>> interoperability and capability to work as a federated system. Therefore
>> the adherence to the defined and deployed protocols is very important.
>> You cannot design a new protocol and expect every service provider to
>> jump wagon the next day. Therefore you will have to continue to accept
>> mail from system speaking the old protocol as well.
> 
> Extending a capability doesn't need the awareness of the service
> providers at all. If hardened Postfix instances meet each other, they
> could agree to use such an extension without even the knowledge of their
> operators. STARTTLS is such an by-case agreement feature.
> 
>> Therefore you gain nothing from the switch because you cannot know if
>> someone speaking the old protocol does it by lazines or other reason or
>> for exploiting its weakness.
> 
> There is no switch. There could be the extension and ... rule tables. If
> you know a group of peers using such an extension, simply set a
> mandatory flag for each of them. If an adversary manipulates the
> negotiation itself, no mail will be transferred. Thanks to the new
> "Uber-Postfix-Extension" (UPE). ;-)

If you need to maintain table, I fail to see where this adds anything
over client certificates checks or simple authentication.

> It's just the same way, you reject unverified or untrusted server
> connections today.

If it is just the same way, why do you need to implement something more?

> That isn't perfectly true. There's a broad hosting service world
> offering quite different SSH implementations, which are out of the
> control of the given single entity. Each users SSH client has to cope
> with such a different set of SSH servers the same way, MTAs and MUAs
> have to contact their peers. Every application with broad use and done
> by many different implementations can be called a "federal system" with
> the obvious need of interoperability.

You don't understand what a federated systems is and how mail delivery
works. I don't think there is much point in discussion this further
since wrong assumptions can only lead to wrong conclusions.

Cheers,
Daniele

Reply via email to