FnEmail-Model

FnEmail — an email system being designed with Event Modeling (Adam Dymitruk's method) before any code is written. This repository contains no code: it is the model, the worked examples, and the rules deciding what belongs in the model.

Current scope: inbound SMTP, HELO only — no ESMTP, no relay, no outbound. Deliberately narrow, so the method is exercised properly before the surface grows.

The author is building this for himself and intends to release it freely — and the project serves a second purpose that explains the first. FnEmail is simultaneously the first worked example for a modeling tool (see the EventModeling repo), and the markdown here is a prototype of generated output. Two operative consequences: a form that is merely tolerable to write by hand but wrong to read is the wrong answer, because the tedium disappears when a tool emits it; and where a relationship cannot be represented in markdown, record the relationship and accept the flat rendering — never abandon the relationship. ⚠️ None of this is license to over-build: the standing instruction (2026-08-07) is to get the basics working first and refine placement later.

Paths are the source; slices are derived. The model document looks like the definition and the walked paths look like instances of it; the flow of authority is the opposite. Held as a working hypothesis, not a settled ruling — the statement is the method repo's layering.md, whose status banner records the demotion from "ruled"; the register row in docs/DECISIONS.md carries the standing change. It is a change of standing, not a correction — the position was accepted and is now held provisionally, which is a different thing from having been wrong (AGENTS.md rule 14).

What FnEmail is

Not normative. This section is orientation for a newcomer. It is not an input to any test, nothing in the model reads it, and the owner changes it without a ruling. It lives in the README because the README already answers what is this repository, and what is the thing being modeled is that same question one step out. What the model claims conformance against is a separate and checkable thing — docs/event-model.md → The normative set. A sentence of prose is not a conformance claim, which is why this one is not asked to be.

FnEmail is an open source, AGPL email server that will be RFC compliant. That includes SMTP for sending and receiving; local users, across multiple domains and multiple accounts; relays, retries and bounces; and IMAP. It will probably end up having its own web-based email client, and spam filtering and prevention will probably work their way in as well. Stated by the owner, 2026-08-09 — and both probablys are his: those two are likely, not committed.

Five repositories

The repos mirror the modeling layers. Canonical ruling record: docs/DECISIONS.md.

Repo	Layer	Holds
EventModeling (public)	method	generic modeling material — the canon primer, layering, path/step form, the three-fates split, rendering, altitude, the parked extensions; the modeling tool it works toward will live in repos of its own
FnEmail-Model — this repo (renamed from FnEmail 2026-08-08, old URLs redirect)	model	the SMTP model, walks, RFCs, SMTP-specific references. Cites the method repo
FnEmail (future)	code	what the model generates — not yet created
ModelingTool-Model (public)	the tool's model	ideas for the modeling tool now, its model later — pre-model, and nothing in it is a decision
event-modeling-research (private)	sources	the mirrored corpus — cite, never reproduce

Method-generic insight flows to EventModeling with FnEmail demoted to example; SMTP specifics stay here (AGENTS.md rule 11). ModelingTool-Model models a tool that implements the method, so it is a consumer of EventModeling rather than a second copy — insight found while modeling the tool still flows up.

Get set up

Clone the repos you need. (Verified working 2026-08-08.)

# 1. The model — this repo. Work is on main; an explore/ branch may be
#    checked out while an exploration runs (AGENTS.md rules 13 and 15).
git clone git@github.com:IvanTheGeek/FnEmail-Model.git /home/ivan/FnEmail-Model

# 2. The method repo — PUBLIC
git clone git@github.com:IvanTheGeek/EventModeling.git /home/ivan/EventModeling

# 3. The research — PRIVATE, needs auth
git clone git@github.com:IvanTheGeek/event-modeling-research.git ~/event-modeling-research

All use SSH. If you get a 404 on the private repo that means not authenticated, not does not exist — it is the confusing failure mode. Generate a key with ssh-keygen -t ed25519, add the .pub half at https://github.com/settings/keys, and check with ssh -T git@github.com.

Verify:

git -C /home/ivan/FnEmail-Model branch --show-current   # → main, or an explore/ branch
ls /home/ivan/FnEmail-Model/docs                        # → DECISIONS.md, event-model.md, paths/, …
ls /home/ivan/EventModeling/docs                        # → layering.md, path-and-step-form.md, …
ls ~/event-modeling-research                            # → research/, extracted/, 6 root files

⚠️ Never copy anything from ~/event-modeling-research into the public repos — AGENTS.md rule 7. Read that repo's extracted/README.md and research/archive/NOTICE.md before quoting from the extracted book or the machine transcripts.

Read in this order

Start with the method layer, then this repo, then the research repo as needed.

File	Repo	Why
README.md, then docs/	EventModeling	The method layer. The canon primer, layering, path/step form, the three-fates split, rendering, altitude, the parked extensions — the canonical statements this repo cites
AGENTS.md	this repo	The working rules — fifteen, each learned the hard way
docs/DECISIONS.md	this repo	What is settled, and must not be re-litigated (AGENTS.md rule 9)
docs/event-model.md	this repo	The model — v0.3, 12 slices, derived: reconciled 2026-08-08 to the active walk on the owner's waiver of the rule-13 hold; its banner lists what stays open
docs/model-altitude.md	this repo	A signpost, not a ruleset — read it to resolve a reference you met somewhere else. Dated notices say where the altitude material went: the gate sequence and the four tiers unadopted 2026-08-10 and parked in the method repo, the nine classified events retired to git, and there is no charter — the normative set in event-model.md and the product statement above are what replaced it. Old filename and section numbers kept, so the references in the walks still land
docs/smtp-path-vs-mailbox.md	this repo	Path versus mailbox — the SMTP reference distinction the walks lean on
docs/paths/	this repo	The walked paths. STEP-FORM.md carries what SMTP adds to the step form; the generic form is the method repo's
docs/diagrams/README.md	this repo	The model as tables — renders on GitHub and the Claude apps
research/METHOD-REFERENCE.md	research	The method as its authors define it, contradictions named rather than smoothed over
research/CORRECTIONS-v0.1.md	research	20 ways the first draft misapplied the method

The working stance (AGENTS.md rule 13): the active WORKING- file is docs/paths/WORKING-helo-direct-single-recipient-v2.md — under walk-through review; nothing is reconciled to it until the prefix drops. The four EXPLORE- files are reasoning records whose headers state what each settled and what stays open; the active exploration is docs/paths/EXPLORE-declaration-vs-status.md — cite it, not event-model.md, for anything it governs (AGENTS.md rule 13).

Where status lives

Status is tracked at its sources, deliberately — a parallel ledger drifts (this README absorbed docs/HANDOFF.md on 2026-08-08 for exactly that reason; its history is in git):

• Working rules — AGENTS.md. Settled decisions — docs/DECISIONS.md.
• What stays open after the model's reconciliation — the ⚠️ banner atop docs/event-model.md.
• Hotspots — docs/event-model.md → Hotspots; the active walk's own Hotspots section is the live ledger for the ones it touches.
• Work only other repos can do — docs/FOLLOW-UPS.md.
• Everything else, including why any current statement is what it is — the commit messages, long on purpose (AGENTS.md rules 6 and 14): documents carry the present, git history carries the why, and ruling-bearing commits are titled for verbatim retrieval with git log -F --grep='<title>'.

License

Code is AGPL-3.0, documents are CC BY-SA 4.0 — the family ruling, registered in docs/DECISIONS.md; the ruling commit carries the reasoning. Applied here:

• The documents — everything this repository currently holds — are licensed CC BY-SA 4.0.
• Code, if this repository ever comes to hold any (checks, scripts), is licensed AGPL-3.0 — the license the future modeling tool and the future FnEmail code repo will take.
• Not relicensed: third-party material included here retains its original license and is not covered by either grant — currently the IETF RFC texts in docs/rfc/, reproduced verbatim under the IETF Trust's own terms (BCP 78). Brief quotations from the method corpus likewise remain their authors' — AGENTS.md rules 2 and 7. Beyond brief quotation, third-party material enters only when its own terms permit redistribution, always attributed, its terms named where it sits.