Wayside Fiber, Part 3: Who Keeps It True

This is Part 3 of Wayside Fiber, a three-part series. Part 1 built the physical plant; Part 2 built the logical one. This part is about the only layer that actually degrades.

The thirty-year cable and the eighteen-month drawing

State the thesis cold: the glass in our corridor will still meet specification in thirty years. The as-built drawings describing it were last fully accurate about eighteen months after cutover, and nothing in the project that built either one is funded to close that gap.

The physical plant and the record of it degrade on different schedules. Only one of them degrades by default.

Every step down on that orange line is a real event from this series. The midspan access at MP 38.1 added a splice location that exists on no original drawing. The duplicate-address incident in Part 2 happened because a record was not consulted, and its cleanup, if undocumented, is itself another step down. None of these are failures of diligence in the moment; each was a competent crew solving the problem in front of them. The record decays through ordinary work, done well, by people with no mandate to write it down. The cable does not need anyone's help to stay true. The record does, and the help has a payroll cost that has to come from somewhere.

Capital builds, operating sustains

Our corridor was built under a federal discretionary grant, the way a great deal of real wayside fiber gets built. The grant was generous and specific: it paid for design, duct, cable, splicing, testing, commissioning, even the as-built drawings as a deliverable. Then it ended, because that is what capital funding does. It buys things, and it is structurally incapable of buying truth over time: not the OTDR re-shoot in year five, not the hour a splicer spends updating the strand table after a repair, not the person who says no to an undocumented change.

All of that lands on the operating budget, the most contested money on any railroad, where "maintain the fiber records" competes directly with visible, urgent work. An operating budget line with no advocate loses that competition every year, quietly, until the night it becomes the reason a four-hour restoration took sixteen. The institutional design problem of the wayside network is exactly this: the asset is capital, its truthfulness is operating, and no one is born owning the second thing. Part of the fix is organizational (name an owner). But part of it is giving that owner something defensible to point at, which is where an unlikely hero enters: the test set from Part 1.

The OTDR baseline is the condition assessment

Railroads and transit agencies increasingly live under formal asset management regimes: transit's Transit Asset Management rule (49 CFR Part 625) with its state-of-good-repair reporting is the sharpest version, and FRA-regulated roads face the same logic through their own asset and capital planning. These frameworks want the same thing: an inventory, a condition rating per asset, and a defensible basis for replacement timing.

Linear buried assets fit this machinery terribly. A vehicle can be inspected, mileage-tracked, and rated by age and wear. A buried cable presents nothing to inspect: fifty miles of glass underground, invisible, with a nominal 30-year book life that is a guess dressed as a number. So buried plant gets condition-rated by age alone, which produces two expensive failure modes: healthy 25-year-old cable scheduled for replacement it does not need, and a corridor rated "good" by birthday while water quietly eats a splice case.

Now reframe what Part 1 already built. At acceptance, every strand was shot bidirectionally at two wavelengths, and every event got a measured loss. That set of traces is not just a commissioning artifact. It is a quantitative, per-event, spatially-resolved condition baseline for the entire asset. Re-shoot the same strands the same way in year five and the comparison is a condition assessment of a rigor almost no other asset class can offer:

Baseline against re-shoot. The splice at MP 38.1 has moved from 0.06 to 0.35 dB: water is in the enclosure, found from a test set, years before it becomes an outage.

A splice that was 0.06 dB and is now 0.35 is water intrusion announcing itself years early. A section whose slope grew at 1550 nm but not 1310 is ground movement stressing the cable. Unchanged traces are equally valuable: they are evidence, not assumption, that the 25-year-old corridor is in a state of good repair and needs no capital. Either way, the answer to "what is the condition of this asset and how do you know" becomes a measurement with a date on it.

The budget reframe follows. A periodic OTDR campaign proposed as testing is a maintenance expense, first thing cut. The same campaign proposed as the condition assessment program for a nine-figure linear asset, feeding the SOGR reporting the agency is federally required to produce, is asset management, defensible in exactly the venue where operating money gets decided. Same trucks, same test sets, different sentence, different outcome.

What sustainment actually costs

A list readers can lift into a budget request. This is what "keeping the corridor true" means as line items:

• Test equipment, owned and calibrated. An OTDR, a light source and power meter pair, an inspection scope, with calibration as a recurring line. Borrowed test sets produce baselines in formats you cannot compare.
• Splice capability, in-house or on retainer: a fusion splicer, closures, trays, and at least two people per territory who can use them at 2 a.m.
• Spares by lead time, not by consumption. You will almost never use the emergency reel of 48-strand cable, matching closures, and FDP pigtails. You stock them because the failure that needs them will not wait a procurement cycle, and the OSP market's lead times are measured in months.
• A pre-executed emergency restoration agreement with a fiber contractor: rates, response times, and mobilization terms signed before the dig-in, so the night of the failure is spent splicing instead of negotiating.
• The OTDR cycle: the re-baseline campaign on a fixed period (five years is a common rhythm), funded as the condition program described above.
• Record maintenance as named labor. Some fraction of a real position, written into a job description, for keeping drawings, strand tables, and IPAM current. Unowned work is optional work; optional work does not survive contact with an operating budget.
• Training, including the unglamorous kind: the next generation of splicers, and cross-training so the addressing record from Part 2 is not one retirement away from folklore.

None of these items is large. All of them are the first things cut when they are framed as overhead, which is why the framing in the previous section matters.

The documentation set

"Keep good records" is advice nobody can execute. A named set of artifacts, each with an owner and an update trigger, is a system. The corridor's set:

Artifact	Contents	Owner	Updated when
Record drawings	Route, depths, vaults, laterals, houses	Engineering	Any plant change
Strand ledger and splice records	The corridor ledger, splice sheets, and FDP port maps shown in Part 1	Signal/comms engineering	Any splice or assignment change
OTDR baselines	Acceptance and periodic traces, both wavelengths, both directions	Comms maintenance	Acceptance, re-shoot cycle, any repair
IPAM / address record	Every address, VLAN, and port assignment	The seam from Part 2: jointly, in writing	Any assignment, before it is made
Config backups	Switch, gateway, and vital comms configurations	Comms maintenance	Any config change, automatically
FRA test records	Required periodic tests on the signal side, per 49 CFR 236	Signal maintenance	On the regulatory cycle

Two things about this table. First, the gaps announce themselves: if you cannot fill in the "owner" column for one of these on your own railroad, that row is currently a rumor, and now you know which one. Second, notice the artifact that spans departments: the address record from Part 2's seam appears here with a joint owner, in writing, because a record with two informal owners is a record with none.

If the set feels like overhead, price it against the night it gets used. A backhoe finds the cable at 2 a.m. somewhere near MP 31. Walk the restoration twice, once with current records and once without:

Restoration step	Records current	Records stale
Locate the fault	OTDR distance mapped to a known route: under an hour	OTDR gives a distance along a route nobody trusts: hours of driving and digging
Identify what is down	Strand table read off in the office	Toning, tracing, and phone calls to whoever might remember
Restore in the right order	Vital pairs spliced first, corridor signalling back early	Splice all 48 and hope, vital back whenever it is back
Confirm restoration	Compare against baseline traces	Argue about whether 0.4 dB "was always like that"

The plant damage is identical in both columns. The outage duration is not, and the difference is purchased entirely in advance, in the unglamorous currency of the table above.

Change control, or: the only mechanism

Everything above still fails without one final piece, because every artifact in that table shares a single failure mode: change without record. The midspan access, the address cleanup, the swapped optic; each was a change, made by someone competent, that made a record false somewhere else.

The only known fix is boring: a documented engineering change process, sized to the railroad. It does not need a board or software. It needs three properties: every change to the plant, the network, or an address is proposed somewhere visible before it happens, recorded in the artifacts it touches when it happens, and checked by someone other than the person who made it. The bench tech staging Chester's replacement recorder under even the lightest version of this process (look up the address before assigning it; log it after) does not create Part 2's incident. The crew at MP 38.1 under this process leaves behind a corridor whose drawings still tell the truth.

Change control has a reputation as bureaucracy, and badly-sized versions earn it. But strip it to its function and it is the same principle this series started with, applied to information: the plant should fail safe. A change process is how a record fails safe: when it is missing, the change stops and asks, instead of silently making the map wrong.

Close: what actually degrades

Run the series backward. The physical plant is a solved problem: glass, duct, splices, and budgets are engineering, and Part 1's corridor will pass every test for decades. The logical plant is a solvable one: schemas, VLANs, and timers are decisions, and Part 2's corridor makes them correctly. Neither of them is what fails.

What degrades is institutional memory: the accurate drawing, the current spreadsheet, the person who knows why tube 4 was never opened, the budget line with an advocate. That layer has no equivalent of a gravity-dropped relay; it does not fail toward safety, it just fades, and every physical and logical failure that follows lands harder because of it. The corridor's real thirty-year project is not keeping the light on. It is keeping the description of the corridor true enough that the night the light goes out, the people holding the record can trust it.

Back to the series introduction, or start again at Part 1: Where It Is.