Re: Recovery metadata and proactive refresh for self-controlled DIDs

Hi Filip,

I agree that for log-based methods such as did, recovery authorization and
state transitions belong entirely within the method's own architecture. The
event log should remain the authoritative source of truth, and I am not
suggesting that resolvers execute recovery logic or that method-specific
state be duplicated in a DID Document.

What I am trying to understand is where a method-agnostic discovery layer
should live, if such a layer is desirable at all.

Recovery appears to involve two distinct concerns:

Recovery execution — proving authorization, rotating keys, and updating
state.
Recovery discovery — determining whether recovery capabilities exist and
how an application can begin interacting with them.

For methods like did:cel, the first concern is already well-addressed by
the method itself.

The second concern seems broader than any individual DID method. Recovery
ecosystems may involve guardians, custodians, recovery providers, wallet
software, archival systems, and future applications that operate outside
the scope of the method.

This raises a survivability question.

Suppose a DID is created today using a particular wallet and recovery
provider. Ten years later, the original wallet may no longer exist, the
provider may have disappeared, and the user may be using completely
different software.

How should a future recovery client discover the existence of recovery
capabilities and determine how to begin interacting with them?

One possible answer is that DID methods should expose this information
themselves.

Another is that discovery belongs entirely in application-layer protocols.

A third possibility is that DID resolution could provide a minimal,
method-agnostic discovery surface, similar in spirit to service discovery,
while leaving all authorization and execution responsibilities to the DID
method.

I am less interested in advocating for a specific mechanism than in
understanding which architectural layer is most appropriate for long-term
portability and survivability.

If the community believes recovery discovery should not be exposed through
DID resolution, I would be interested in understanding what alternative
mechanism would allow a future client to bootstrap recovery across diverse
DID methods without requiring method-specific integrations for each
recovery architecture.

-- Amir Hameed Mir

On Wed, 10 Jun 2026 at 04:54, Filip Kolarik <filip26@gmail.com> wrote:

> Hi Amir,
>
> My view is that recovery is inherently DID-method specific rather than
> something that should be standardized or exposed at the DID Document level.
> In practice, different methods have very different trust and lifecycle
> assumptions, and recovery logic sits squarely in that space. From that
> perspective, putting recovery metadata into the DID Document doesn’t really
> help interoperability and is not meaningful outside the method. So I’d lean
> toward keeping recovery out of a DID Document and handling it through
> method-specific mechanisms [1].
>
> This topic is important to me, as it needs to be defined for the did:cel
> method, which democratizes infrastructure and provides fully decentralized
> control over DIDs without dependency on blockchain or any central storage
> mechanism.
>
> Curious if anyone has concrete cases where exposing recovery data in the
> DID Document actually adds a value.
>
> Thanks for starting the discussion!
>
> Best regards,
> Filip
> https://www.linkedin.com/in/filipkolarik/
>
> [1] https://github.com/w3c-ccg/did-cel-spec/issues/19
>
>
>
>
> On Wed, Jun 10, 2026 at 1:39 PM Amir Hameed <amsaalegal@gmail.com> wrote:
>
>> Hi all,
>>
>> I've been working on a DID Recovery specification draft and wanted to get
>> some feedback on a few architectural questions before I go much further.
>>
>> The model combines threshold social recovery, guardian-assisted recovery,
>> commitment-based verification of recovery shares, and proactive refresh of
>> recovery material over time.
>>
>> At the moment, recovery information is represented separately from the
>> controller key material and includes things like recovery methods, guardian
>> information, thresholds, commitments, and recovery epochs. The prototype
>> also supports refreshing recovery shares without changing the underlying
>> DID, allowing recovery material to evolve while preserving the recovery
>> policy.
>>
>> The part I'm struggling with is where recovery information should live
>> from an interoperability perspective.
>>
>> Should recovery metadata be represented directly in a DID Document, or
>> should it remain external and be discovered through service endpoints or
>> some other mechanism? If recovery state changes over time, does it make
>> sense to expose any of that through DID resolution, or should it remain
>> entirely wallet/application specific?
>>
>> I'm also curious whether there is value in standardizing recovery
>> concepts independently of any particular DID method. Many ecosystems seem
>> to be converging on similar ideas (social recovery, guardians, threshold
>> schemes, multi-device recovery, etc.), but implementations tend to be
>> isolated from one another.
>>
>> Another area I'm exploring is proactive refresh. In the current design,
>> recovery shares can be rotated and a new recovery epoch established without
>> changing the DID itself. Has anyone worked on similar approaches, and if
>> so, how did you handle auditability, discovery, or interoperability around
>> those changes?
>>
>> I'd appreciate any feedback, references to related work, or pointers to
>> previous discussions that touch on recovery, guardians, threshold
>> cryptography, or recovery event tracking in DID systems.
>>
>> Thanks,
>>
>> Amir Hameed
>>
>

Received on Wednesday, 10 June 2026 12:34:22 UTC