Re: [Announcement] Procedural Memory Knowledge Representation (PM-KR) Community Group — Call for Participation

Hi Manu,
Thank you so much for this response—it means a lot coming from you, and 
the CBOR-LD + rdf-canon connections are exactly the kind of alignment I 
was hoping PM-KR could explore with the W3C community!
## CBOR-LD and Compression Tables
I'm really excited you pointed to CBOR-LD! The compression table 
approach there (term dictionaries, compact representation) aligns 
beautifully with what PM-KR is trying to generalize.
In K3D (the reference implementation motivating PM-KR), we're seeing 
~70% compression via symlink-style composition—for example, the 
Character Galaxy compresses from 87.7 MB (static font payloads) to 26.3 
MB (procedural Bézier programs + references). The key insight is that 
the "compression table" isn't just term mappings; it's *canonical 
procedural forms* that can be referenced, composed, and executed.
Your mention of "generalized solution to build the compression tables" 
is exactly what PM-KR aims to study: how do we construct, maintain, and 
canonicalize these procedural reference tables across domains (not just 
linked data terms, but visual forms, spatial structures, mathematical 
symbols, etc.)?
## RDF Canonicalization and Procedural C14N
The rdf-canon connection is fascinating! The idea of "transcluded 
graphs" built from procedural c14n is precisely the pattern we're seeing:
- **Current rdf-canon**: Canonicalize RDF graphs for digital signatures
- **Procedural c14n (PM-KR direction)**: Canonicalize *procedural 
compositions* so you can sign over symlink-style references
For Verifiable Credentials with large, repetitive structures (as you 
mentioned), this could mean:
- Store canonical procedures once (e.g., "standard credential schema" as 
a procedure)
- VCs reference those procedures via stable identifiers
- Sign over the *composition* (procedure references + instance data), 
not the fully expanded payload
This preserves both compression and cryptographic integrity—exactly what 
large-scale VC deployments need.
## Alignment Questions (Lightweight, When Bandwidth Allows)
Even if you can't participate directly, I'd love your occasional 
guidance on a few alignment questions as PM-KR develops:
1. **CBOR-LD compatibility**: Should PM-KR's compression table mechanism 
be designed as an *extension* of CBOR-LD's term dictionary approach, or 
as a parallel/complementary system? Where do the boundaries make sense?
2. **RDF-canon integration**: For procedural c14n, should we aim to 
produce outputs that rdf-canon can *also* canonicalize (layered 
approach), or explore a unified canonicalization that handles both 
triples and procedural compositions?
3. **Verifiable Credentials use case**: Are there specific VC patterns 
(repetitive schemas, large credential graphs) where a pilot integration 
with PM-KR compression would be most valuable? We'd love to ground this 
in real-world data if Digital Bazaar or the VC community has 
representative examples.
## Next Steps
I'll make sure PM-KR's charter and early deliverables explicitly reference:
- CBOR-LD (compression tables, term dictionaries)
- RDF Canonicalization (c14n for digital signatures)
- Verifiable Credentials (large, repetitive credential graphs as a key 
use case)
And I'll watch the JSON-LD WG discussions closely to ensure PM-KR 
complements rather than duplicates the excellent work already happening 
there.
## Thank You
Your "cheering from the sidelines" means the world, Manu.
Knowing that PM-KR's direction aligns with CBOR-LD and rdf-canon—and 
that you see potential value in generalized compression tables and 
procedural c14n—gives me confidence we're on the right track.
Even lightweight input (occasional email, GitHub issue comment, "hey, 
you might want to look at X") would be incredibly valuable as we move 
forward. No pressure, just an open invitation whenever bandwidth allows. 🙂
Looking forward to building this in a way that strengthens the whole W3C 
linked data ecosystem!
Best regards,
Daniel Ramos
Founder, Knowledge3D Project
Proposed Chair, PM-KR Community Group
daniel@echosystems.ai
https://github.com/danielcamposramos/Knowledge3D
https://www.w3.org/community/pm-kr/


On 2/20/26 6:20 PM, Manu Sporny wrote:
> On Fri, Feb 20, 2026 at 4:10 PM Daniel Ramos<capitain_jack@yahoo.com> wrote:
>> **Interoperability and Compression**: PM-KR complements RDF/JSON-LD by providing procedural execution semantics and compression through symlink-style composition.
>> ### 🤖 For AI & Machine Learning Communities
>> - **Compression without information loss**: Symlink-style composition reduces duplication while preserving semantic fidelity
> Current work that is happening in that general space:
>
> https://w3c.github.io/cbor-ld/#abstract
>
> https://lists.w3.org/Archives/Public/public-json-ld-wg/2020Jul/att-0004/Introduction_to_CBOR-LD.pdf
>
> We'd love to see a more generalized solution to build the compression
> tables (your focus).
>
>> ### 🕸️ For Knowledge Graph Communities
>> **Graph Compression and Execution**: PM-KR complements property graph and RDF graph approaches by introducing procedural canonicalization—storing graph patterns as reusable procedures rather than duplicating structure.
> Related, but not the same:
>
> https://www.w3.org/TR/rdf-canon/
>
> This work could benefit from different, efficient c14n mechanisms
> where we could digitally sign over "transcluded graphs" built from
> procedural c14n. Might be useful for very large and repetitive W3C
> Verifiable Credentials.
>
> Exciting work, wish I had the bandwidth to more directly participate
> -- will be cheering from the sidelines. :)
>
> -- manu
>