Documenting Address Space: The IP Plan That Survives You
Every network has a memory of its address space. The only question is where that memory lives: in documents anyone can read, or in one person's head. The head version works surprisingly well — right up until the head goes on vacation, changes jobs, or simply forgets why 10.1.20.7 must never, ever be touched. Then the network starts making expensive guesses. Address documentation is the working memory of the network moved somewhere durable. The standard it must meet is blunt: it has to outlive its author. Not survive until the next reorg. Outlive. You are not writing notes for yourself; you are writing the letter your successor opens on their worst day.
This guide is the capstone of the address-management story. It covers what exactly to write down, in what shape, and — the part everyone skips — how to keep it true without heroics. The answer is smaller than you fear. There are three documents, each with a template you can copy from this page. One reconciliation habit acts as the heartbeat. And a four-question test tells you honestly whether what you have written is documentation or decoration.
In this guide
- The three documents
- Document one: the address plan
- Document two: the assignment register
- Document three: the conventions page
- Keeping it true
- The handover test
- Format and home
- When the documents pay for themselves
- The build order
The three documents
Address documentation wants to be exactly three artifacts — not one giant sheet, not a wiki sprawl. Each answers a different question at a different altitude. The address plan is the strategic view: which subnets exist and why — the map of the territory. The assignment register is the tactical view: within each subnet, which individual address belongs to what — the census of the citizens. And the conventions page is the decoder ring. It supplies the rules and meanings that make both other documents — and every address on the network — interpretable without a guide. Three documents, three altitudes, and together they hold everything a stranger needs. If you read the IPAM basics guide, you will recognize the first two as the mature forms of its registers. This guide grows them into a document set built for handover.
Document one: the address plan
One row per subnet, covering the entire space you own — including the parts that are deliberately empty. The plan answers "what is this range, and where does the next thing go?" It is the document you consult before creating anything. Its design decisions — the sizing, the function-based carve, the headroom — come from subnet planning. This table is where those decisions are written down so they keep operating after you stop explaining them:
| Range | Site | Purpose | VLAN | Gateway | DHCP portion | Growth notes |
|---|---|---|---|---|---|---|
| 10.1.10.0/24 | HQ | Users | 10 | 10.1.10.1 | .50–.199 | .11.0/24 reserved — pairs into /23 |
| 10.1.20.0/24 | HQ | Servers | 20 | 10.1.20.1 | none | Statics .20–.49; .21.0/24 reserved |
| 10.1.50.0/24 | HQ | Guest / IoT | 50 | 10.1.50.1 | .20–.229 | Split IoT to .51 when count justifies |
| 10.1.60–.250 | HQ | Unallocated | — | — | — | Future functions — do not squat |
| 10.2.0.0/16 | Branch | Site block | — | — | — | Same carve as HQ |
Notice what earns a row: the unallocated ranges, listed with the same seriousness as live subnets. An empty range that is documented is a growth reserve; an empty range that is undocumented is a squatting opportunity. And notice the growth-notes column — it records intent, which is the one thing no scanner can ever rediscover for you. A tool can re-learn what exists; only the plan remembers what was meant.
Document two: the assignment register
The register descends into each subnet: one section (or sheet, or tool view) per subnet, one row per static address and DHCP reservation. Dynamic leases stay out — the DHCP server documents those automatically, and copying them by hand is a chore with an expiry date. The register is for the addresses a human chose:
| Address | Hostname | Type | Purpose / why | Owner | Assigned | Last verified |
|---|---|---|---|---|---|---|
| 10.1.20.7 | lic01 | Static | License server — ERP vendor binds licenses to this exact address; changing it means a vendor call and a reissue fee | IT | (date) | (date) |
| 10.1.20.11 | fs01 | Static | File server | IT | (date) | (date) |
| 10.1.20.31 | prn-acct | Reservation | Accounting printer — reserved so scan-to-folder never breaks | IT | (date) | (date) |
Feed it automatically wherever you can. Discovery scans and DHCP reservation lists can populate and verify most columns — what answers at each address, its MAC, its hostname. A register wired to those sources stays close to reality with almost no typing. The Network Discovery & Asset Inventory guides cover building that feed. But note which column no feed can fill: the why.
Give each subnet's section a two-line header while you are at it. Repeat the subnet's plan row (so the reader never juggles two documents mid-lookup). Add a one-liner on its character. "All-DHCP, no statics live here" is itself valuable information. It tells the 2 AM reader that anything static-looking on this subnet is an intruder by definition. An explicitly empty section and a missing section read very differently to a stranger: one says "nothing here, and we checked". The other says "nobody ever looked."
Tip: the "why" column is the difference between a register and a list. "10.1.20.7 — lic01" tells your successor what is there; the why tells them the license server's address is load-bearing and expensive to change. Every hard-won, invisible constraint you record there is one 2 AM mistake a future admin — possibly future you — does not make. Write the why while you still remember it; it is the only column with a deadline.
Document three: the conventions page
The shortest document does the most quiet work. The conventions page states the rules that give every address its meaning. Those are the octet scheme, the gateway rule, and the reserved-band logic from subnet planning. With those rules, any address, anywhere on the network, decodes on sight. The conventions page is one page, and it looks like this:
| Convention | Rule |
|---|---|
| Address shape | 10.SITE.FUNCTION.HOST — second octet is the site, third is the function |
| Sites | 1 = HQ, 2 = branch; new sites take the next /16 |
| Functions | 10 users, 20 servers, 30 management, 40 voice, 50 guest/IoT; odd numbers between are growth reserves |
| Gateway | Always .1, every subnet, no exceptions |
| Host bands | .2–.19 infrastructure, .20–.49 statics/reservations, .50–.199 DHCP, .200+ reserved |
| VLANs | VLAN number = function octet (subnet .20 = VLAN 20) |
| Never touch | Unallocated ranges in the plan are reserved on purpose — allocate only via the plan |
The payoff compounds daily. This page takes five minutes to absorb. With it absorbed, a reader translates 10.2.30.4 into "branch, management subnet, an early static, probably a switch" without opening either other document. The conventions page is what makes the whole address space self-describing; the other two documents fill in the specifics, but this one teaches the language they are written in.
One honest wrinkle for inherited networks: the network you have probably breaks whatever rules you write, because it grew before the rules existed. Do not let that stop you, and do not pretend. Write the conventions you want going forward, then add a short exceptions list at the bottom of the page. For example: "10.1.20.88 is a server in the wrong band; predates the plan; migrate at next rebuild." A documented exception is a known quirk with a retirement plan. An undocumented one is a trap wearing a normal address. Over the years the exceptions list shrinks, and watching it shrink is one of the quiet satisfactions of the job.
Keeping it true
Now the part that separates documentation that works from documentation that lies politely. Documents drift the moment they are finished, because they record intentions while the network keeps moving. The countermeasure is the reconciliation loop — the scheduled comparison of register against discovered reality described in IPAM basics — running as this documentation's heartbeat. Diff a discovery scan against the register on a schedule. Every mismatch lands in a small queue — the ghost entries, the unrecorded squatters. Working that queue weekly is what keeps the register a source of truth instead of a source of nostalgia. A register with a heartbeat can be trusted at 2 AM. One without is a rumor with formatting.
Just as important: edits ride inside the workflows, never alongside them. Documentation fails when updating it is a separate virtuous act — virtue loses to tickets every single day. So bolt each edit to the change that causes it. Start with the static-assignment intake rule: the register row is written before the device is configured. The register is where you get the address from. Next comes the subnet-addition checklist (plan row first, then VLAN, scope, rules, monitoring). Then comes the decommission outtake, the forgotten twin. When a system retires, its rows are marked released the same hour. Undead entries are exactly how "free" addresses stop being findable. Do that, and the documents update themselves as a side effect of ordinary work. Then two light reviews catch what slips through. A quarterly skim takes twenty minutes: does each subnet's section still smell right, and is the mismatch queue empty? An annual deep pass is where you re-verify the register's aging rows and prune what died quietly. That is the entire maintenance burden: workflows carry the edits, the loop catches the leaks, two calendar entries catch the rest.
The handover test
How do you know whether what you have written is any good? Not by rereading it — you wrote it, so it makes sense to you; that is precisely the trap. The quality bar is the handover test. Picture a competent stranger — a new hire with solid networking basics and zero knowledge of your organization. Could they sit down with only these three documents and answer the four questions that actual operations ask?
- 1. What is this address? Hand them
10.1.20.7from a firewall log. They should reach "license server, do not touch, vendor-bound" in under a minute, via conventions and register. - 2. Where do I put a new server? The plan names the subnet; the conventions name the band; the register names the next free address. If any hop requires asking a human, that hop is your gap.
- 3. What is actually free in the server subnet? Free-and-verified, not free-by-rumor — which tests whether the reconciliation loop is real and the "last verified" column is recent.
- 4. Which ranges must never be touched? The growth reserves, the vendor-bound statics, the ranges with landmines — the question whose wrong answer causes the most expensive incidents.
Run the audit on your own documents — honestly, pretending amnesia, or better, by handing a colleague the four questions cold. Every stumble marks a missing row, an unwritten why, or a convention that lives only in your head. And note where this test points: address documentation belongs in the break-glass pack. That is the sealed-envelope set of documents an organization opens when the only person who knew things is unreachable. If you are the whole IT department, that pack is not paranoia, it is professionalism; the IT Department of One guides treat the survival kit properly.
Format and home
Format questions generate strong opinions, so here is the principle that settles most of them: queryable beats pretty. The address plan gets consulted under stress and filtered by need — "everything at the branch," "every DHCP portion." Structured tables answer those instantly, whether they live in a spreadsheet, a wiki's tables, or a proper IPAM tool. A gorgeous network diagram with addresses drawn on it is a fine companion. Diagrams have their own craft, covered in the Network Documentation guides. But as the system of record, a diagram fails the moment you need to sort, search, or diff it against a scan. Documentation you cannot query is documentation you will eventually stop consulting.
Two more decisions matter more than the tool choice. One authoritative home. The plan might exist in five copies — the wiki, two spreadsheets, a PDF in email, whatever the vendor was sent. The moment it does, each copy drifts independently. Readers pick their favorite, and now the documentation disagrees with itself. Pick one home; everything else is an export, clearly labeled as such, regenerated on demand and never edited. Exports are also your resilience. Keep a periodic plain-file snapshot (CSV does fine) stored with your backups — and in that break-glass pack. That means the day the wiki or the IPAM tool is itself the casualty, the address plan is still in hand. And set access deliberately: make it readable by everyone who might ever need a lookup. That includes helpdesk, since "what is this address" is half their job. But make it editable by the few who own the workflows. Wide reading keeps the documents alive and used; narrow writing keeps them true. A document only insiders can read fails the handover test by construction.
Warning: never let the address documentation live only inside the network it documents. The night you need it most is the night the network is down. If the plan sits solely on a file server whose address you would need the plan to find, you have built a locked box with the key inside. The periodic export to the break-glass pack, kept reachable without the network, is the escape hatch.
When the documents pay for themselves
Two composite stories, both of which happen on every network eventually — the only variable is whether the documents exist when they do. The outage: a Tuesday, 2 AM, the monitoring system pages about a device flooding the network from 10.1.20.44. Without a register, the first hour of the incident is spent answering "what is that address?" — sweeping, tracing switch ports, waking people who might remember. With the register, the same question takes thirty seconds: reservation, accounting printer, there is its MAC and its why. Unplug it and go back to bed. Total incident time is shorter than this paragraph's backstory. Same failure, same fix — the entire difference was whether the network's memory was written down. The audit: a cyber-insurance questionnaire, or a security review, asks for your network segments, their purposes, and the addressing of critical systems. That request either triggers a two-week archaeology project or a ten-minute export from documents that were already true. Teams that have lived the second version never go back; being boringly ready is a genuine pleasure. And there is a third payoff nobody budgets for: onboarding. A new admin handed these three documents is doing useful, safe work on the address space in their first week. That is because the documents are the training, written once and delivered to every future colleague for free.
The build order
If you are starting from nothing, do not start alphabetically — start in dependency order, smallest first:
| Order | Document | Effort | Seed it from | It stays true via |
|---|---|---|---|---|
| 1 | Conventions page | An afternoon | Decisions you make once (or rules you excavate from the existing network's habits) | Rarely changes; review annually |
| 2 | Address plan | A day | Router configs, VLANs, DHCP scopes — what exists — plus the conventions for what should | Subnet add/retire checklists |
| 3 | Assignment register | Grows forever | Discovery scans, DHCP reservations, then the squatter-hunt | Intake, outtake, weekly reconcile |
Conventions come first because they cost the least and shape everything after. The plan comes second because it frames every register section. The register comes last and forever, because it is never finished — it just gets truer. On an inherited, undocumented network, the raw material for all three comes out of a structured week of discovery — the discovery sprint ends with exactly the subnet and device registers this document set grows from.
Where to go next
Three documents — plan, register, conventions — kept true by workflows and a heartbeat, tested against the four questions a stranger would ask. That is address documentation that outlives its author. And it is the capstone of everything in this topic. It brings together the discipline from IPAM basics, the design from subnet planning, and the watchfulness from DHCP scope monitoring. All of it is written down where it can save someone who never met you. The rest of the IP Address Management & Subnet Monitoring guides are the natural companions. When you are ready to document more than addresses, the Network Documentation topic covers the wider ecosystem this joins.
Put this into practice with SysPeep Monitor
Documents stay honest only when facts keep flowing into them. SysPeep Monitor supplies the facts. Auto-discovery keeps a live inventory of the devices on every subnet. Its reports export cleanly, so your register always has fresh reality to reconcile against. The free trial installs on a Windows machine in minutes.
Download Free TrialLearn About SysPeep