I became involved with Project Xanadu in the early 90s, during its ending days. Although its hypertext system was before its time, one feature that particularly struck me was the Xanadu Club system, which allowed for read and write permissions to be passed among both users and clubs (groups) in a fascinating recursive manner that was quite different from anything else I'd ever seen before (or since!).
The problem? The architecture was centralized.
Welcome, Christopher!On Wed, Sep 24, 2025 at 6:29 PM Christopher Allen <Christ...@lifewithalacrity.com> wrote:I became involved with Project Xanadu in the early 90s, during its ending days. Although its hypertext system was before its time, one feature that particularly struck me was the Xanadu Club system, which allowed for read and write permissions to be passed among both users and clubs (groups) in a fascinating recursive manner that was quite different from anything else I'd ever seen before (or since!).While the overlap isn't perfect, much of what was done by the Club system can now be done with RBAC systems that include role inheritance. The Club system mostly predates the current preference for Capability-driven intuitions. That said, there are a number of things Clubs support that don't have a graceful implementation with capabilities.The problem? The architecture was centralized.The Xanadu implementation at the time was centralized. The underlying architecture was not. In retrospect, I suspect that some of MarkM's later work on eventual consistency may have had its origins in efforts to make certain parts of the Xanadu architecture asynchronously distributed.
Perhaps he will comment.
Cryptography is a fine way to implement, but it doesn't (in and of itself) address the need for eventual convergence of independently made updates to the access rules.
Sounds like maybe you've got something interesting to add to the discussion!Jonathan
--
You received this message because you are subscribed to the Google Groups "friam" group.
To unsubscribe from this group and stop receiving emails from it, send an email to friam+un...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/friam/CAAP%3D3QPEEJSAug%3D1ryLqS_sgyyBi5nc9JEhD%3DkVvPP2%2BQeJrDA%40mail.gmail.com.
On Wed, Sep 24, 2025 at 10:02 PM Jonathan S. Shapiro <jonathan....@gmail.com> wrote:Welcome, Christopher!On Wed, Sep 24, 2025 at 6:29 PM Christopher Allen <Christ...@lifewithalacrity.com> wrote:I became involved with Project Xanadu in the early 90s, during its ending days. Although its hypertext system was before its time, one feature that particularly struck me was the Xanadu Club system, which allowed for read and write permissions to be passed among both users and clubs (groups) in a fascinating recursive manner that was quite different from anything else I'd ever seen before (or since!).While the overlap isn't perfect, much of what was done by the Club system can now be done with RBAC systems that include role inheritance. The Club system mostly predates the current preference for Capability-driven intuitions. That said, there are a number of things Clubs support that don't have a graceful implementation with capabilities.The problem? The architecture was centralized.The Xanadu implementation at the time was centralized. The underlying architecture was not. In retrospect, I suspect that some of MarkM's later work on eventual consistency may have had its origins in efforts to make certain parts of the Xanadu architecture asynchronously distributed.<self indulgent. skip if you don't care>Indeed. I started working on Xanadu in 1976 or so, just before the RSA papers. Absorbing Ted's ambitions for Xanadu -- which do resemble the cypherpunk perspective 10 years later -- I tried to solve the needed distributed systems problems myself. I could not. This led to four things:
- fascination with Actors, for reasons that are now obvious ;)
- leaving Xanadu to take a job at Datapoint, the inventor of the first good local area network (Arcnet) and a decentralized OS that ran across large numbers of Datapoint's tiny desktop computers -- 64K memory altogether. I went there for several reasons, but mainly because it seemed to me the best place to learn about distributed systems. That OS was fascinating and unlike anything I've seen since.
- went from Datapoint to Xerox PARC, which just so happened to become the new best place to learn about distributed systems. This was more accident than intentional. I wanted to go there mostly because of Smalltalk.
- joining the early cypherpunk movement.
- by the time I left PARC to return to Xanadu, taking Dean with me, I felt I was ready.
But the work that you all know me for, ocaps with decentralized secure communicating event loops, does not itself involve eventual consistency.
- Xanadu did have strong elements of eventual-consistency, though I do not recall ever explaining things in those terms.
- Dean and I did important eventual-consistency work at the "Agoric" company of the 1990s under the term "available objects" and written in Joule. But little of that saw the light of day.
- The one part that made it into E was the Lamport Cell (only inspired by Lamport), that was really cool, but was not redone in E's successors. All that eventual-consistency logic is mostly subsumed many years later by CRDTs.
The actual history is of course much messier than my tidy story above.</self indulgent. skip if you don't care>Perhaps he will comment.Christopher's breakthrough here is that it is not just decentralized, it is infrastructure-less. This is an extreme level of protection that I never imagined would be possible for something like this.Cryptography is a fine way to implement, but it doesn't (in and of itself) address the need for eventual convergence of independently made updates to the access rules.Christopher will be explicit about what kind of consistency can be achieved under his constraints. It is not eventual consistency, but IMO it is still quite valuable. I am excited!--Sounds like maybe you've got something interesting to add to the discussion!Jonathan
You received this message because you are subscribed to the Google Groups "friam" group.
To unsubscribe from this group and stop receiving emails from it, send an email to friam+un...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/friam/CAAP%3D3QPEEJSAug%3D1ryLqS_sgyyBi5nc9JEhD%3DkVvPP2%2BQeJrDA%40mail.gmail.com.
--Cheers,
--MarkM
To view this discussion visit https://groups.google.com/d/msgid/friam/CAK5yZYgAyxkOACW%3D3LLPwnR%2BFLNcBrzfBJfwWKwrymJOCpPG1g%40mail.gmail.com.
To view this discussion visit https://groups.google.com/d/msgid/friam/CAFybx9%3DXbK%3DO-LgnxtmPgdJMCn-1gfP3LkGx%3DK2Ooy7-%2BxQEUQ%40mail.gmail.com.
I expect to be busy Wednesday morning. Is there any chance the
presentation will be recorded so I can watch it later?
Christopher's breakthrough here is that it is not just decentralized, it is infrastructure-less. This is an extreme level of protection that I never imagined would be possible for something like this.
Cryptography is a fine way to implement, but it doesn't (in and of itself) address the need for eventual convergence of independently made updates to the access rules.Christopher will be explicit about what kind of consistency can be achieved under his constraints. It is not eventual consistency, but IMO it is still quite valuable. I am excited!
That being said, there are a lot of interesting possible given scriptless scripts / adapter signatures (see https://raw.githubusercontent.com/LLFourn/one-time-VES/master/main.pdf & https://eprint.iacr.org/2024/1809.pdf) - basically anything that can be expressed as a series of zk-proofs can be used as an adapter signature. Other new cryptography such as VDF (Verifiable Delay Functions) offer other possibilities.