[ros-users] frame_id in headers

Ken Conley kwc at willowgarage.com
Fri Oct 7 00:01:33 UTC 2011

This may be a case of telephone game (i.e. a misinterpretation), or perhaps
there's yet-another mistake involved, but the way I would put it is: "making
Header a first-class concept in ROS was a mistake."

Header is more of a "TF Header".  This has created couplings we wish we
didn't have in the client library (roscpp, rospy) code.  Prior to ROS 1.0 we
tried to cleanup the main ROS client library code to have no robotics in it;
this is the one case that was too difficult to pull out due to the large
amount of code that utilizes the Header data structure.

There is currently no alternative.  At some future point in time, where it's
worth the cost of being 'clean', we can undo this, but it is currently the
case that the costs of undoing it to end users far outweigh the benefit.
Our current resolution to this is that we migrated Header to be
'std_msgs/Header' and we will continue to try and de-specialize its role as
much as backwards-compatibility allows.

 - Ken

On Thu, Oct 6, 2011 at 4:39 PM, Geoffrey Biggs <geoffrey.biggs at aist.go.jp>wrote:

> Morning all,
> I've heard occasionally from various people that putting frame_id in the
> Header is considered a design mistake that we're now stuck with, at least
> until a new version where the API can be broken. Can anyone involved comment
> on why it's considered a mistake, and what the preferred alternative is?
> Geoff
> ______________________________**_________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/**listinfo/ros-users<https://code.ros.org/mailman/listinfo/ros-users>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20111006/96973338/attachment-0004.html>

More information about the ros-users mailing list