Community consensus report on DNS WG structures at IETF

Findings The small team found some clear points of consensus points within the collected opinions. These findings were later distilled into recommendations ().

A DNSDISPATCH mechanism would be beneficial for deciding where and how new work should be formed.
- Working groups can then concentrate on the work they are chartered for.
- Followers know where to follow new works of interest.
- A downside is a potential slow down of new work, and an increase in agenda time.
Creating two groups, one for operations and one for protocol development, would be helpful.
- One would concentrate on operations and hopefully streamline the process to get from drafts to RFCs.
- One would concentrate on longer term protocol development efforts, potentially in a higher-volume discussion.
- A downside discussed is that some people would need to attend and participate in both groups anyway. Though this is clear for some IETF participants, there were indications it doesn’t apply to everyone. Some may also be able to concentrate fully on one, and merely watch the other.
No structure can solve the “human problems”.
- It is still up to the area directors and chairs to deal with disagreements of all kinds.
- This includes how and where work is handled in more nuanced cases.
- WG chairs need to be supported in handling these situations.
- WG chairs MUST coordinate within and between groups and discuss DNS@IETF wide current topics of concern with each other and their ADs.
Narrow chartered working groups are necessary for more challenging development problems
- DELEG and ADD being two examples, with DELEG being an especially agreed-upon example of an that needed a separated, dedicated working group.
We did not receive feedback indicating that the other DNS groups not mentioned here, like DNSSD and REGEXT, need structural modifications.

Recommendations Based on the findings above, and extrapolating information from discussions to derive a suitable path forward, the DNS@IETF small team recommends that the area directors considering the following advice:

Create a new DNSPROT (DNS Protocol) or similar group for working on protocol development and maintenance.
- This group should have a fairly wide charter that tasks it with work on the DNS protocol itself.
- Things requiring special processing rules likely belong in DNSPROT
- Documentation about protocol semantics should be in DNSPROT
Create a new DNSDEP (DNS Deployment) or similar group for working on protocol deployment and operational concerns. [Really need a better new name]
- This group should have a fairly wide charter that tasks it with work that doesn’t require special processing rules, needs algorithms or other simple IANA actions, or are BCPs that document existing behaviours.
- Examples include algorithm assignments, IANA actions, BCPs, etc.
- “How you use the protocol”
- Alg roles, bcps, split horizon, zone cut to nowhere
Work toward closing DNSOP in order to properly signal the change
- Keep it open and functional until all current work is finished
- Some work already in progress in DNSOP could move to DNSPROT or DNSDEP where work would continue, at the discretion of the authors and chairs
Create a DNSDISPATCH group for providing guidance to authors about where new DNS work should be conducted.
- To avoid introducing delays and agenda constraints, this group should conduct its work almost entirely over a mailing list with only difficult cases requiring interim or, worst case, in-person meeting time. Ideally, in-person meetings should be rare.
- DNSDISPATCH can recommend dispatching work to dnsprot/dnsdep/AD-sponsored/another-WG/BOF/ISE.
- DNSDISPATCH may decline to provide a recommendation for documents that are not within scope, for example.
- Chairs of the group need to be strict in enforcing and carrying out its objective.
- The DNSDISPATCH group will not prioritize work within the other groups, and its dispatch decisions cannot result in automatic adoption.
- A significant portion of submissions to DNSDISPTACH can likely be handled quickly and efficiently.
- The DNSDISPATCH chairs should require that documents clearly articulate the problem space and proposed solution before consideration.
- The DNS directorate is a resource available to the DNSDISPATCH working group, just as it is available to other working groups.
- The dispatch group might use a pool of willing shepherds to assist the chairs and authors with process related help for incoming documents.
- The dispatch group will make informed recommendations to document authors about where to take their work
  - The output of a dispatch discussion should include a short shepherd write up (perhaps a paragraph in length)
    - Light weight write ups that are sent to the mailing list for archiving. This should not require datatracker changes.
    - DNSDISPATCH chairs should create a light template text as a boiler plate to be used by most cases.
  - DNS WGs MAY require in their charter that new work first gets a dispatch suggestion before consideration in their WG.
  - After a dispatch, document authors are encouraged to follow the recommendation and approach the WG chairs with a follow-on request (including but not limited to adoption requests).
  - Each group will continue to follow its own processes for formal adoption.
- The chairs of the DNSDISPATCH group should work closely with the chairs of the other groups. They may need to work together for handling more difficult topics and to collaborate on advice or questions for the DNSDISPATCH WG participants.
Group management is expected to be significantly different in each of these groups.
- With an effective split in functionality, it allows each group to have different forms of execution, meeting, progression, and publication requirement strategies.
- For example, some groups may require running code, while others may not.
Documents may occasionally (rarely we hope) need to move after being dispatched when the problem scope changes during its development and refinement.
- For example, problems that become large may need to move to a new group.
- Sometimes, however, the decision will be wrong but might as well stay in the current group.
- The area director and WG chairs will need to handle this (rare) problem on a case by case basis.

Example Dispatch Scenarios The small team recognized that some examples might be helpful in better understanding how the envisioned DNSDISPATCH group might process incoming work. As such, we came up with three example scenarios to highlight how we envision some workflows might happen.

Maxwell Coulomb writes a document that describes a new way that DNS can be used by DHCP clients. They take this document to DNSDISPATCH where, after some discussion including references to other discussions in DHCP working groups, the chairs post a recommendation drawn from consensus to the list saying that in their opinion the best DNS working group for this document would be DNSDEP. Maxwell then approaches the DNSDEP chairs by sending a message to the chairs that includes a link to the DNSDISPATCH recommendation. The chairs review and decide that this is a good candidate document for DNSDEP to consider and send a request for comment to the DNSDEP mailing list.
Marie Ampère writes a document that describes a new protocol for encoding video into linked, short ASCII messages, including examples of how this allows video to be published in the DNS. They take this document to DNSDISPATCH where, after some discussion, the chairs post a recommendation that this is not a good fit for any DNS working group since it does not really represent DNS-specific work. Thus, the chairs decline to provide a recommendation.
Marmaduke Nxdomain writes a document in response to some operational problems that have been discussed in another forum, proposing some changes to DNS best practices to avoid such failures. After some discussion, including references to presentations and related observations surfaced in a recent DNS-OARC meeting, the chairs decide that this is a good candidate document for DNSDEP but that the document would benefit from some restructuring and rewriting first so that the substantive issues can be better considered in the working group. The chairs solicit a volunteer shepherd to help Marmaduke with the next steps. The shepherd helps Marmaduke update the text and later discuss the document with the DNSDEP chairs, including a reference to the DISDISPATCH recommendation.

Suggestions that achieved no or only fairly rough consensus

Always requiring running code.
- Running code before adoption definitely did not have consensus.
- Running code before publication had generally rough consensus.
- Based on this, we believe each group will need to make their own decision on this matter as suggested above.
BCP documentation is an open question about where best to develop them.
- Some believe operational groups like DNS-OARC should drive BCP development.
- There is rough consensus that publication of BCPs should remain in the IETF.
- It may be that interim meetings held in conjunction with external conferences would be a good idea.