Process for deploying new BGP attributes (experimentally or otherwise)

Following our recent attempts to experiment with potential use of BGP attributes,
I would like to understand better the desired process for using new BGP attributes.
As mentioned earlier, we investigate currently one usage of BGP attributes for improving BGP security - specifically, adoption of RPKI. I also have some future ideas on other ways BGP attributes may be useful to improve BGP security; and surely other applications, security or otherwise, may exist. So I think we need an acceptable process to do it.

Indeed, while we aborted that experiment, I doubt that the right conclusion is that one should simply never use a new BGP attribute… And I even hope that the community would want us to complete our research, to see if the particular approach we evaluate may indeed be beneficial. I think that these sentiments were also expressed by many members of this community.


  • Is there some document discussing the desired (best practice) process? [if not, maybe such document should be written?]
  • We used unassigned attribute. Should one always assign an attribute before using it? That’s a possibility, but has some `costs’ - and may yet fail to prevent problems to some non-conforming networks.
  • Is there an agreed-upon list of the forums and mailing lists on which one should warn in advance about such planned announcements, and the details that should be included?

Thanks, Amir