Skip to main content

Minutes IETF125: wimse: Tue 03:30
minutes-125-wimse-202603170330-00

Meeting Minutes Workload Identity in Multi System Environments (wimse) WG
Date and time 2026-03-17 03:30
Title Minutes IETF125: wimse: Tue 03:30
State Active
Other versions markdown
Last updated 2026-03-19

minutes-125-wimse-202603170330-00

WIMSE @ IETF125

Chairs: Pieter Kasselman, Justin Richer
Notetakers: Micke Nordin

Agenda & Chair Updates (5min): Chairs

Slides:
https://datatracker.ietf.org/meeting/125/materials/slides-125-wimse-welcome-and-chair-updates-00

Justin and Pieter open the meeting remotely, Orie is on site stand in.
The Note Well is well noted. WIMSE is two years old now, and we should
start to finish up. Where do we go next after we have wrapped up the
current work items? Seven active drafts in the WG, none is finished,
although one is in last call. A very nice illustrative cartoon curtesy
of Fox (13), shows how one draft has been split up in to four.

WIMSE Workload Credentials (8min): Brian, Joe, Arndt, Yaron, Yaroslav

Draft:
https://www.ietf.org/archive/id/draft-ietf-wimse-workload-creds-00.html

Slides:
https://datatracker.ietf.org/meeting/125/materials/slides-125-wimse-wimse-workload-credentials-00

Brian presents. The WG decided to split service to service draft in to
components, this presentation focuses on the credential part. x509 is
the basis for this. not much work has been done, but
draft-ietf-wimse-identifier-02 has been referenced.

Brian Campbell referred to “credential” as the identifier in a
certificate, so Kathleen asked a question to clarify. They consider the
public/private key pairs (he did didn’t say this initially likely due to
jet lag) PLUS the identifier for the workload of the token or X.509
certificate and the binding as the credential. Kathleen offered to
review the draft and help on language or to point out where Kathleen
found it confusing or a different use of terms - Brian welcomes the
reviews to ensure the document is useful to the broader community.

Workload Proof Tokens (8min): Brian

Work done:

  • Clarify treatment of "jti" claim
  • fix Example WPT Claims to match the Example WPT
  • inline ABNF for workload proof token
  • Add audience security considerations

Please open a PR if you feel like you should be acknowledged in the
document.

HTTPSig (8min): Yaron

Draft:
https://www.ietf.org/archive/id/draft-ietf-wimse-http-signature-02.html

Slides:
https://datatracker.ietf.org/meeting/125/materials/slides-125-wimse-http-signatures-00

One major change since the last document, close to WGLC. We want to make
sure we have an audience under the signature, so as to make sure it can
not be reused somewhere where it was not intended.

Problems are that some environments rewrite the URL, and that it can
happen in many places and hetrogeneously across environments. this is
addressed by WIMSE-Audience HTTP Header.

Brian thinks that this should indeed be considered a major change and
that the new header seems awkward. Brian thinks signing a part of the
authority piece is important, but the general feeling is that this is
the wrong aproach.

From chat (Yaron Scheffer)
For the record, I would have preferred to add an explicit "derived
component" for the Audience, but that would not be allowed by 9421. For
the record, I would have preferred to add an explicit "derived
component" for the Audience, but that would not be allowed by 9421.

WIMSE Architecture (8min): Joe

Draft: https://datatracker.ietf.org/doc/draft-ietf-wimse-arch/
Slides:
https://datatracker.ietf.org/meeting/125/materials/slides-125-wimse-wimse-architecture-update-00

Identifier has now been moved to the Identifier document. Workload and
Workload Instance has been defined and the Bootstrapping section has
been revised. Discussion welcomed.

Henk asks if it is better if it should be clairified that an instances
relies on the capabilites of the thing it runs on.

From the chat (Kathleen)

I think if attestation were to get tied in (in a different draft) than
Henk's comment would be needed. I think it can be out-of-scope for now
and added later when the combinations happen - it can be in layers

From chat (Brian) Kathleen's comment ^ seems about right

Yaroslav: There are two ways to do this:

From chat (Richard Wilhem): I think that a Workload Instance need not be
"running"
It could have occurred in the past
(definitions are hard)

From Chat (Flemming): I agree the workload instance may be "paused",
etc., but once it's terminated it is no more. That doesn't mean the
identifier associated with it ceases to exist at that point in time.

  • For those deployments,where instance location host matters that can
    encode it the path of identity so you could say VM something
    something or you have Kubernetes, something, that would identify
    that but that obviously would beyour deployment specificthing.
  • Another way is attestation would attest to a platform where you host
    it and then you could have all wonderful cryptographic evidence
    associated with that

From chat (Antoine Fressanaut): a noob's remark on terminology: As I am
approaching WIMSE, having the terminology defined consistently in the
Architecture document (Which I consider my entry point as a newcomer to
WIMSE) would be nice

Workload Identifier (8min): Yaroslav

Draft:
https://datatracker.ietf.org/meeting/125/materials/slides-125-wimse-http-signatures-00

Slides:
https://datatracker.ietf.org/meeting/125/materials/slides-125-wimse-wimse-workload-proof-token-00.pdf

  • Defined workload identifier origin
  • Removed ovelap with s2s credentials work
  • Added URI requirements
  • Added WIMSE URI scheme definition
  • Spelled out that identifiers may apply to workload instances and
    logical workloads

Do we need a separate registry for workload schemes or should the IANA
URI scheme registry suffice? The authors thinks no separate registry is
needed.

Do we want additional wimse:// requirement or leave it flexible -
authors think no

We need to align terminology across all documents, preferably in a
single place that can be referenced from other places.

Flemming thinks the drafts are in a good place and cudos to the authors.

Peter (Liu) likes the URI scheme and thinks the problem is the public
domain is trying to define some kind of global, parsing mechanism that
must consider the gaps between the global resolution versus local
resolution and consider the global may use DNS or similar
infrastructures so the top level, identifier could be the trust domains
of each differentand administrative domains so i think this could be a
really good design to help the global private administrative donors.

Identity Practices WGLC (5min): Arndt

Draft:
Slides:
https://datatracker.ietf.org/meeting/125/materials/slides-125-wimse-wimse-workload-identity-practices-00

Clarified identity vs credentials.
Explain what federation actually produces
Consistent terminlology across credential types.
Tighten SPIFFE terminology.

Issue 69 is about Differentiate Workload Identity Document where Cloud
providers offer Instance Metadata Endpoints which issues two distinct
documents.

The document has been in WGLC for a while so please review.

The chairs ask who can commit to reviewing the document. Flemming offers
to review as well as Kathleen who will do another review when it is
done.

AOB (10min): Chairs

Are we nearing the end here, should we go to WGLC all around, or do we
need more work done?

Brian thinks that it is mostly true, but there still remains some work
to be done.

Kathleen asks newer participiants to review the documents, as it is a
good way to contribute in a meaningfull way.

Please let the chairs know if Interim meetings needs to be scheduled to
discuss any new topic.