w3c
Make conformance statements concretely testable
@msporny suggested we need to lock down to a specific serialization for the inputs and algorithms.
This will apply for sections 4 - https://www.w3.org/TR/did-resolution/#resolving and 5 https://www.w3.org/TR/did-resolution/#dereferencing.
For further context on the issue please see https://www.w3.org/2025/10/09-did-minutes.html#8d71
Also Manu had suggested one potential normative statement could be:
Normative statement: Implementations must provide results that are serializable to JSON.
We could try something along these normative statements:
For the "software library" conformance class, in section 4 algorithms state something along the lines of:
Implementations MUST receive inputs and produce outputs that are serializable to JSON.
For the "network-based" conformance class:
Implementations MUST be a
You can do this for the last part of that statement:
Section [[[#bindings-https]]]
You can also do variations of this in ReSpec:
[SPEC-REFERENCE] [[[SPEC-REFERENCE-TITLE]] [[[#SPEC-SECTION-NAME]] <a data-cite="EXTERNAL-SPEC#SECTION-REFERENCE">some external spec section reference</a>
[=auto-link to term=] <-- alias for <a>term</a>
All documented on https://respec.org/docs/
This was discussed during the #did meeting on 30 October 2025.
View the transcript
Make conformance statements concretely testable #213
<ottomorac> w3c/
Wip: Have to make sure that all the needed parties can be at the special call.
<ottomorac> Issue deals with defining conformance classes for DID Resolution, which is required for resolving other issues. Conformance classes should be defined in terms of inputs and outputs, Manu has proposed some wording.
manu: Reference to section 4 -- have to return something in one conformance format. So must be JSON. For DID resolution, it is concretely testable, right? Could make it so input parameters must be capable of being serializable to JSON.
… Single sentence in both cases...should be easy...
Wip: Yes, but should be added, so needs two be clarified. Other part is adding in the conformance class. Are we aligned on that class name? We need to conformance classes.
manu: We can name them anything -- doesn't really matter. Would be nice to resolve. "Software Library Resolver" and "Network-based Resolver". Software library deals with internal data structs which is OK if serializable. For the Network -- should be clear, use HTTP binding.
Wip: That sounds great. No one is assigned. I'll do it or ottomorac. Sound fine -- let's get something in.
ottomorac: You can have it Wip.
