[ros-users] ROS & DDS

[Ken Conley]

This has been an interesting discussion to follow. Sorry to chime in
late (I'm lacking time to write e-mails), but some quick thoughts:

A single middleware solution was good for ROS in the early days, as it
kept things simple, and thus more usable. As ROS scales to more use
cases, that same simplicity potentially hurts usability, because the
homogeneity lacks the structure to simplify configuration. Also, more
structure would also create a more clear robot stacks, where each
stack may require different architectures (e.g., how the navigation
stack and the PR2 ethercat driver each have custom plugin
architectures).

So, in a discussion of DDS and otherwise, is it possibly the case that
issues of backwards compatibility/etc... can be side-stepped by
looking at specific use cases (e.g., multi-robot, offboard
monitoring), and building more specific architectures/libraries/tools
around those (at the risk of losing homogeneity)? If these solutions
are successful, they can be expanded to other areas.

This is similar to past discussions like:

 * tf and the transform tree (i.e., avoid distributed topics, and
instead have a shared memory protocol or similar)
 * integrating video streaming protocols

 - Ken


On Thu, Feb 20, 2014 at 3:45 PM, Brian Gerkey <gerkey at osrfoundation.org> wrote:
> On Thu, Feb 20, 2014 at 3:26 PM, Aravind Sundaresan
> <asundaresan at gmail.com> wrote:
>> We are using ROS in a DARPA program where we need the communication between
>> the master and nodes as well as the messaging to be secure. Is there anybody
>> else who finds this important? Does DDS provide secure communications?
>
> You're not alone in wanting secure communications for robotics.  We
> would be remiss in a major rewrite to completely ignore security
> (which is not to say that we must implement security mechanisms, but
> we'd better at least end up with a solid story about our approach to
> security, with clear guidance to users).
>
> As I understand it, there's a proposed extension to the DDS spec,
> currently under review, to add security.  Here's a relevant
> presentation:
>
> http://www.slideshare.net/GerardoPardo/dds-security-standard
>
> But given that DDS is apparently used in so many mission-critical
> applications, surely there's some encryption happening, perhaps in
> non-standard vendor-specific ways that vary by implementation?  Or
> perhaps they're using DDS like we use ROS, where you're expected
> secure your network.
>
> brian.
> _______________________________________________
> ros-users mailing list
> ros-users at lists.ros.org
> http://lists.ros.org/mailman/listinfo/ros-users