[zeromq-dev] Encryption (OpenSSL/TLS)

Pieter Hintjens ph at imatix.com
Fri Oct 1 19:21:07 CEST 2010


On Fri, Oct 1, 2010 at 7:13 PM, Oliver Smith <oliver at kfs.org> wrote:

> No, it is just the wrong way to use 0MQ as internet-scale fabric...

There is IMO only one issue with defining a tls:// transport.  It
cannot be aware of messaging patterns and so cannot span devices.  It
must necessarily be a single-hop encryption just as tcp:// is
necessarily a single-hop protocol.

This doesn't however break any use cases anyone has identified.

While VPNs and per-message encryption seem fundamentally flawed.  The
notion of asking users to install OpenVPN in order to get a secure
link is... horrid.  And per-message encryption leaves the problem of
key exchange unsolved, and makes applications aware of encryption, and
leaves them option to replay attacks.

So unless someone has a specific explanation why a tls:// transport
would not solve the use cases we have, or somehow break 0MQ
scalability, it seems to be the correct solution.

My earlier idea of using a pair of devices is just a hack to avoid
putting this into 0MQ core, it's essentially the same solution.  It
would be a good proof-of-concept IMO.

I'm tempted to dig up the OpenSSL interface code from Xitami...

-Pieter



More information about the zeromq-dev mailing list