Re: [Isms] Discussion: Architecture direction for ISMS

[Eric Rescorla <[e-mail]>]

Robert Story <[e-mail]> writes:

> On Wed, 13 Apr 2005 08:28:22 -0400 Martin wrote:
> MS> Why does the reverse communication require any different configuration,
>
> Again, I'm using the specific example of ssh in its current form. To
> communicate with a peer:
>
> 1) generate my private/public keys
> 2) configure peer with my public key
> 3) connect to host (and peer host key is save for future reference)
>
> This is the existing infrastructure that I am referring to. Note that while I
> do have a key for the host, it is only used to verify that I'm talking to the
> same host the next time I connect.
>
> To allow the host to connect back to my machine autonomously, I would have to:
>
> 4) generate a private/public key on the peer (possibly could re-use host key)
> 5) configure my host with the peer's public key
> 6) install/configure/run ssh server
>
> The point is that these steps are extra, and not part of the configuration of
> the existing infrastructure.

In SSL, at least, this problem has already been attacked in the
context of FTP, which involves callbacks on the data channel.
Basically, you do session resumption but the party that does
the active open (what one would think of as a TCP client)
acts as the SSL server. See draft-murray-auth-ftp-ssl-16.txt
(though the text isn't as clear as it could be).

Note that this assumes you want to tear down the TCP
connection. If you don't, then there's no problem, of course.
Incidentally, with DTLS you can just leave the DTLS association
up.

-Ekr



_______________________________________________
Isms mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/isms