The initial version of NESTS encloses all entities within a single filesystem tree 1. That may be sufficient in the initial stages (executable documentation, experimentation, proof of concept, development, etc.) but it will never be able to support a very large number of accounts and identities, especially given the fundamental limitations of filesystem sizes.
There are also good reasons to maximize localization and to minimize centralization:
A fairly simple approach preserving the current simplicity would be to distribute the accounts, identities, currencies and the specific namespaces enclosing these across an arbitrarily large collection of nodes 2, the individual entities being accessible (when necessary) from other nodes via SSH 8, 9.
The global namespace map would require a very small fraction of the storage required by the entities themselves and it might be practical to hold a duplicate copy on every node (replicated continually using RSync) but an alternative would be to provide a smaller network of dedicated entity-mapping nodes 10.
There may be several ways to achieve this, but one easily implemented approach would be to use SSH for all inter-hub activities, using request-confirmation handshaking.
The process might be successfully limited to the following categories:
parent-child signalling - in particular the propagtion of the .enabled flag state from a parent namespace in one hub to a child namespace in another hub.
In this case, an instruction can be issued immediately by the parent namespace's stewards to the child namespace's hub supervisory code to change the state of child namespace's .enabled flag. Unfortunately is cannot be guaranteed to get through immediately (or at all), the parent's hub would have to continue re-issuing the instruction until a confirmation response is received. To ensure idempotency, a unique instruction ID can be generated for each such request.
Transactions on or between accounts.
Operations upon entities (including creation, modification, deletion and suspension).
However, the initial version of NESTS will be confined to a single node (or hub) and all that matters in the short term is that a simple approach is available to scale up in due course.
This does not necessarily mean within one single filesystem. Most *nix filesystems allow for the mounting of a filesystem upon a node within another filesystem, and these need not always be of the same type.
One option to include *nix filesystems (of many types) withing a NESTS entity tree would be to use NFS. However, even a relatively small number of NFS mounts will quickly overwhelm the system (see note here for example).
The term "node" here means an installation, at whatever location, running on any compatible operating system, and serving communities using whatever natural language can be accommodate using UTF-8 encoded name strings.
The term "power" here means simply the ability to enact decisions.
A noun.
A verb.
The entities, as always, include:
The global stewardship is required to provide an overarching policy, ethos and identity (VSM System 5 and to oversee and arbitrate upon the allocation of namespace names.
The burden need not be unmanageable if, as seems reasonable, the overwhelming majority of transactions were to occure within rather than between nodes.
Whereas entities are currently identified by FIP (Full Internal Path) or FPH (Full Path Hash), inter-node access would require only sufficient detail to enable SSH access:
These are broadly analoguous to /etc/hosts and DNS respectively.
Version 0.2 - 2021-10-12 16:00 - Copyright © 2021 John Waters