Daniel sent us this one — he's got a new rental apartment, fiber to the ONT in the living room, all the networking gear sitting there, but his workstation needs to be in another room. There's existing conduit with a coax run, and an electrician suggested looping back through the communications cabinet by the front door to reach the target room. Daniel's already set up a stopgap OM3 fiber run between switches, but now he's wondering: should he pull fiber through the wall permanently, or just go with Ethernet? And the deeper question — can you even terminate fiber optic cable in the field without ISP-grade fusion splicing gear, or is that whole idea a nonstarter?
This is exactly the kind of constraint-based puzzle I love. You've got a real topology, real physics, a rental budget, and an electrician who's well-meaning but clearly not a network person. Let's map this out. The ONT and OPNsense router live in the living room. The workstation is twenty to thirty meters away in another room. There's conduit between them, but it already has coax in it. The electrician's proposal — go from the living room back to the comms cabinet, then loop back to the target room — that's physically possible, but it doubles the cable length and adds at least two more bend points. Before we even get to fiber versus copper, that loopback idea is worth scrutinizing.
Right, because if there's a clear conduit path directly from the living room to the target room, you pull straight through. You don't detour to the comms cabinet just because it's there. Every extra meter of cable and every extra bend is a failure point waiting to happen. But Daniel mentioned the conduit already has coax in it, so the direct path might be tight. That's constraint number one.
Constraint number two is the rental factor. He's not invested in this apartment. Anything he puts in the walls stays in the walls. So the question isn't just "what's the best cable" — it's "what's the best cable that doesn't cost me a fortune to install and abandon." Let's start with the physics, because that's where most people get tripped up. OM3 fiber has a minimum bend radius of about seven and a half millimeters for long-term static installation, according to Cable Matters' 2025 guidelines. But during the actual pull — the dynamic bend radius — you need at least fifteen millimeters. Cat6a copper can tolerate about twenty-five millimeters without signal degradation, and honestly, copper is way more forgiving if you accidentally kink it during installation.
Here's the thing about fiber — if you exceed that bend radius even once during the pull, you've permanently damaged the glass. It's not like copper where you get a bit more crosstalk and the link still negotiates at a lower speed. Fiber just goes dark. You've now got a very expensive piece of glass in your wall that does nothing.
I want to pause on that, because "goes dark" sounds dramatic, but it's literally what happens. There's no graceful degradation with fiber. Copper is analog in its failure modes — you bend it too hard, you get packet loss, you get retransmits, the link might drop from ten gig to five gig to one gig, and you can still limp along. Fiber is digital in its failure — light gets through, or it doesn't. The attenuation crosses a threshold, and the transceiver on the other end simply loses sync. No warning, no partial service.
That's a really important distinction. With copper, a bad pull might give you a flaky link that you can troubleshoot and maybe live with temporarily. With fiber, a bad pull gives you a dead cable and a do-over. The loopback route the electrician proposed makes this worse. Two one-eighty-degree turns. Fiber can handle those IF the bend radius is respected at every point — but during a pull through conduit that already has coax in it, maintaining that radius is genuinely difficult. The coax is heavy, it's already in there, and as you pull fiber alongside it, the friction alone can cause micro-bends. OM3 fiber has a maximum tensile load of fifty Newtons during installation. Exceed that, and you get permanent attenuation — the glass stretches microscopically and never recovers.
For anyone who doesn't think in physics units, that's about the force of holding a five-kilogram weight against gravity. It's not much. If your fish tape snags and you give it a good yank, you've just ruined the cable.
Think about the last time you pulled anything through a wall. Think about that moment where the fish tape catches on something, and your instinct is to give it a sharp tug. That's easily a hundred Newtons. You've just doubled the tensile limit of your fiber cable, and you won't know it until you plug everything in and get no link light. That's the hidden cost of fiber installation — the failure is silent and delayed.
The bend radius question has a clear answer: fiber can handle the turns IF you're careful, but it's inherently riskier than copper during installation, especially in conduit that's already partially occupied. Now let's talk termination, because this is where Daniel's real question lives. He said he was always under the impression that terminating fiber requires ISP-grade gear. That's a common belief, and it used to be true. But the Fiber Optic Association — the FOA — confirms that field termination is absolutely possible with mechanical splice connectors. No epoxy, no polishing. Kits like the Corning UniCam or the AFL FastConnect run between two hundred and four hundred dollars for a full kit.
Let me be clear about what that means. A mechanical splice connector is a pre-polished fiber stub inside the connector body. You cleave the field fiber flat, insert it into the connector, and a mechanical clamp or gel aligns it with the stub. You don't need a fusion splicer, which costs five thousand dollars and up. You need a cleaver, a stripper, some cleaning wipes, and the connectors themselves. It's entirely doable on a kitchen table.
Reliable for a first-timer? That's a different question. The success rate for someone doing their first mechanical termination is maybe sixty to seventy percent. And here's the catch — when a termination fails, you can't just recrimp it like Ethernet. You have to cut off the connector, re-cleave the fiber, and try again. Each failed attempt eats about twenty to thirty millimeters of cable. On a thirty-meter run, you've got margin. But on a run that's been cut to exact length, two or three failed terminations and suddenly your cable is too short.
Compare that to Ethernet. A twenty-dollar crimper, a bag of RJ45 ends, and after one YouTube tutorial you're hitting a ninety-nine percent success rate. If you mess up, you snip the end off and try again — it costs you a few cents and thirty seconds. The skill floor for copper termination is basically "can you follow color-coded instructions." The skill floor for mechanical fiber termination is "can you cleave glass cleanly without introducing a chip or angle that ruins the insertion loss.
I want to dwell on that cleaving step, because it's the part that surprises everyone. Cleaving isn't cutting — it's scoring and snapping. You use a precision tool that makes a microscopic scratch on the glass, then applies tension to create a perfectly flat break. If your cleaver is dull, or if there's a speck of dust on the fiber, or if your hand shakes slightly, you get a chip, a lip, or an angled face. Any of those defects scatter light at the connection point. Your insertion loss goes from an acceptable 0.3 decibels to 2 or 3 decibels, and suddenly your link budget is blown.
Here's the thing — you can't see the defect. Not with the naked eye. You don't know the termination is bad until you put a light source and power meter on it, or you plug it into actual equipment and it doesn't link up. With Ethernet, a bad crimp is visible. You can see that the wires aren't seated, or the jacket isn't in the strain relief. With fiber, you're flying blind without test equipment.
This is where the electrician's confusion is instructive. He suggested creating another ONT outlet, which reveals a common misconception — the fiber mystique. The belief that anything involving glass requires ISP-level equipment and training. In reality, mechanical connectors have democratized fiber termination. Anyone with steady hands and a three-hundred-dollar kit can do it. But democratization doesn't mean it's wise for every situation.
Let's put numbers on this. Pre-terminated thirty-meter OM3 patch cable, off the shelf — twenty-five to forty dollars. A field termination kit plus bulk OM3 cable plus connectors — two hundred fifty to four hundred dollars, and that's before you factor in the learning curve and potential wasted connectors. Pre-terminated thirty-meter Cat6a — fifteen to twenty dollars. Cat6a bulk cable plus a crimper and ends — fifty to eighty dollars. The fiber premium is five to ten times the cost for zero performance gain at this distance.
Zero performance gain. That's the part that needs to land. OM3 supports ten gigabit per second up to three hundred meters. Cat6a supports ten gigabit per second up to one hundred meters. Daniel's run is twenty to thirty meters, or sixty if he does the loopback. Both cables deliver identical throughput. The only advantage fiber gives is future twenty-five or forty gigabit speeds — which no home user needs, and likely won't need before this rental lease ends.
This is the forward-thinking trap. It's seductive. You think, "I'm pulling cable anyway, why not pull the one that can do forty gig?" But you're in a rental. You're not going to be there when twenty-five gig home networking becomes relevant. And even if you were — twenty-five gig over copper, 25Gbase-T, is expected to become affordable around late twenty twenty-seven or twenty twenty-eight. By the time you actually need more than ten gig in a home, copper might have caught up.
I want to push back on the "future-proofing" argument more broadly, because I see it everywhere in home networking. People pull OM4 in their walls for a ten-gig link today because someday they might want forty gig. But let's be honest about what "someday" means. Forty gig networking in a home requires forty-gig switches, forty-gig NICs, and a workload that can actually generate forty gig of traffic. That's not streaming 4K video. That's not even streaming eight uncompressed 4K video streams simultaneously. Real-time uncompressed 8K video editing from a NAS? Multi-user VR with raw sensor data? These are enterprise workloads, not apartment workloads.
Even if you are that one person doing uncompressed 8K editing in a rental apartment — which, respect — you've already got a working OM3 patch cable between switches. You can upgrade that to OM4 or OM5 later without touching the walls. The in-wall run doesn't need to be the future-proof link. It just needs to be the reliable link.
Let's talk about the stopgap solution Daniel already set up, because it's actually the smartest thing in this whole scenario. He ran OM3 between two switches — it's not in the wall, it's presumably along the baseboard or under a rug, and it works. That proves the link is viable. It's also repurposable. When he moves out, he unplugs it and takes it with him. That's the rental-friendly fiber deployment. The question is whether to make it permanent and in-wall.
The answer, for this specific scenario, leans hard toward no. Here's why. If you're going to pull cable through conduit in a rental, you want the installation to be as close to foolproof as possible. Copper gives you that. The bend radius is more forgiving. The termination is trivial. The cost is lower. And if you abandon it when you move out, you're abandoning maybe twenty dollars of cable and five dollars of connectors. With fiber, you're abandoning a more expensive cable plus whatever you spent on the termination kit that you might never use again.
I want to address the loopback route specifically, because the electrician wasn't entirely wrong — he was solving a different problem. If the direct conduit path between the living room and the target room is blocked or too tight with the coax already in there, the loopback through the comms cabinet is a fallback. But even then, a single continuous pull is better than two segments joined at the cabinet. Every connector, every patch panel, every transition point adds insertion loss and another thing that can fail. If you must do the loopback, pull one continuous cable the whole way.
If you're pulling sixty meters instead of thirty, you're now within the range where fiber's distance advantage starts to matter — except it still doesn't, because Cat6a does ten gig at up to a hundred meters. You'd need to be over a hundred meters before fiber's performance advantage becomes real. Sixty meters is still comfortably inside copper's envelope.
Let me give you a concrete analogy here. Think of copper and fiber like two delivery trucks. Copper is a reliable box truck — it carries ten gig of cargo, it handles potholes and tight turns without complaining, and its range is a hundred kilometers. Fiber is a high-speed train — it can carry forty gig of cargo, it's faster in theory, but it needs smooth rails, gentle curves, and a specialized loading dock at each end. For a thirty-kilometer trip, both vehicles arrive at the same time with the same cargo. You're paying for the train but getting no benefit from it.
In this scenario, the "loading dock" is the termination process we've been describing. The box truck backs up to a standard loading dock — your RJ45 crimper. The train needs a custom platform — your mechanical splice kit. For the same thirty-kilometer trip.
There's one scenario where fiber wins at short distances, and it's worth naming because it might apply to someone else listening. If you need electrical isolation — say you're running cable between buildings, or you're in an area with gnarly ground potential differences — fiber is the answer. Glass doesn't conduct electricity. No ground loops, no surge propagation, no lightning risk. But for a single apartment run under fifty meters? Copper is the correct answer.
I want to give a real cautionary tale on ground potential differences, because people underestimate this. I once saw a setup where someone ran copper Ethernet between two buildings on a farm. Different electrical panels, different ground rods. The ground potential difference was about forty volts. That forty volts traveled up the Ethernet shield and fried every piece of equipment on both ends within a week. Fiber would have prevented that entirely. But inside a single apartment? You're on one electrical panel, one ground. The isolation benefit of fiber is irrelevant.
Let's also talk about what happens if you try to pull fiber through conduit that already has coax. The coax is heavy, it's stiff, and as you pull fiber alongside it, the coax can press the fiber against the conduit wall at bend points. That's how you get micro-bends — tiny deformations in the glass that cause signal loss without visibly breaking the cable. Copper can shrug this off.
Micro-bends are insidious because they're cumulative. One micro-bend might add 0.1 decibels of loss. Ten of them adds 1 decibel. Your total link budget for a ten-gig OM3 link is maybe 2 to 3 decibels, depending on the transceivers. You don't have much margin. A conduit pull alongside existing coax is basically a micro-bend factory.
This is where the tensile load rating becomes critical. Fifty Newtons for OM3. When you're pulling through conduit with existing cable, the friction increases significantly. You're not pulling through an empty tube — you're pulling alongside something that's already occupying space and creating drag. It's very easy to exceed fifty Newtons without realizing it, especially if the conduit has any tight turns.
Let's build two budgets. Option A: Cat6a direct pull. Thirty meters of pre-terminated shielded Cat6a, about twenty dollars. Maybe you buy it pre-terminated and pull it with a pulling grip to protect the connector. If you're doing bulk cable and terminating yourself, add a crimper and ends — call it sixty-five dollars total and about two hours of labor. Success probability: extremely high.
Option B: OM3 field-terminated through the loopback route. Sixty meters of bulk OM3, about forty dollars for the cable alone. Mechanical termination kit, three hundred dollars. Connectors, another thirty. You're at three hundred seventy dollars and probably four hours of labor, with a thirty percent chance you'll need to re-terminate at least one end. And if you kink the cable during the pull, you're starting over.
The math is not subtle. For a rental, for this distance, for this use case — copper wins on cost, wins on installation complexity, wins on reliability, and delivers identical performance.
I want to complicate this slightly, because there's a version of this story where fiber is the right call even at short distances, and it's worth naming so we're not just saying "copper always wins." If Daniel owned this apartment, and he knew he'd be there for ten years, and he wanted to run 25-gig or 40-gig in five years — then pulling pre-terminated OM3 with pull-proof connectors through empty conduit would be a defensible choice. The key differences are ownership, time horizon, empty conduit, and pre-terminated cable. None of those apply here.
That's the key. The rental changes everything. In a house you own, you can afford the learning curve. You can pull fiber, mess it up, pull it out, try again. You can invest in the termination kit knowing you'll use it for multiple rooms over multiple years. In a rental, you get one shot, and if it fails, you've spent hundreds of dollars and an afternoon for nothing.
Now, I want to be fair to fiber. If Daniel wants to learn fiber termination — and I completely understand that impulse, it's a useful skill — the right approach is not to make your first attempt in a rental wall. Buy the kit, buy some bulk OM3, sit at your desk, and practice terminating a dozen connectors. Measure your insertion loss. Get your success rate above ninety percent. Then, when you're confident, you can install fiber anywhere. The existing OM3 patch cable between switches is actually a perfect learning tool — swap in different SFP plus modules, experiment with different wavelengths, understand how the link behaves before you commit to in-wall installation.
That patch cable is doing double duty. It's providing a working network link right now, and it's a sandbox for learning fiber without the commitment. Smart move on Daniel's part.
Here's a fun fact for the learners: SFP plus modules are cheap now. You can get 10G SR modules for fifteen to twenty dollars each on the used market. So your practice setup — a couple of switches with SFP plus cages, that OM3 patch cable, and some cheap transceivers — that's a sub-hundred-dollar fiber lab. You can learn everything about link budgets, wavelengths, and connector types without ever touching a wall. Then when you do eventually need in-wall fiber, you'll know exactly what you're doing.
Let's crystallize the actionable advice. For this specific scenario — rental apartment, twenty to thirty meters, existing conduit with coax, no future-proofing requirement — buy a pre-terminated Cat6a shielded cable and pull it. It's cheaper, it's easier, and it performs identically. If the direct conduit path is available, pull straight from the living room to the target room. If it's blocked, the loopback is a fallback, but pull one continuous cable, not two segments.
If you must go fiber — because you want to learn, or you're planning for speeds beyond ten gig — buy pre-terminated OM3 with pull-proof connectors. LC duplex with a pulling eye built in. Do not attempt field termination as a first-timer on a cable that's going into a wall. Practice on a bench first.
The one scenario where fiber is the unambiguous winner? If your run exceeds a hundred meters — not your case — or if you need electrical isolation between buildings — also not your case. For a single apartment run under fifty meters, copper is the correct answer.
The electrician's loopback idea — only use it if the direct path is blocked. Even then, a single continuous pull beats two segments joined at the comms cabinet. Every connector is a potential failure point and a source of signal loss.
I think there's a bigger question lurking here, and it's about where home networking is heading. Right now, ten gigabit is the ceiling that matters for almost everyone. But 25Gbase-T is on the horizon — expected to become affordable around late twenty twenty-seven or twenty twenty-eight. When that happens, will copper finally hit a wall that forces home users to fiber? Or will ten gigabit remain sufficient for the next decade of home use?
I suspect ten gig will be plenty for a long time. The applications that saturate a ten-gig link in a home are... Uncompressed eight-K video editing? But for the vast majority of users, even power users, ten gig is headroom they'll never fully use. By the time they need more, they'll be in a different apartment with different conduit.
Even the eight-K editing argument is thin. Most eight-K workflows use compressed codecs. ProRes 422 HQ for 8K runs about 400 megabytes per second — that's 3.You'd need multiple simultaneous streams to saturate ten gig. How many people are running multiple uncompressed 8K streams in a rental apartment?
The bandwidth demands of home users grow, but they grow slowly. We went from 100 megabit to 1 gigabit over about fifteen years. From 1 gigabit to 10 gigabit will probably take another decade or more. And by then, the hardware will be cheaper, the standards will be settled, and Daniel will be in a different apartment.
That's the rental reality. You're solving for the lease you have, not the house you might own someday. Daniel's existing OM3 patch cable between switches is already doing the job. If he wants wired internet in that room right now, he's got it. The question was whether to make it permanent, and the answer is: not with fiber, not at this cost, not in this apartment.
Unless you really want an excuse to buy a fiber termination kit. In which case, I respect the hustle, but don't pretend it's about practicality.
Now: Hilbert's daily fun fact.
Hilbert: In the high medieval period, a phantom island called Saint Brendan's Isle was charted in the Atlantic based on an Irish monk's voyage account. Cartographers kept drawing it for over six centuries. Chemical analysis of medieval inks from those maps shows high concentrations of iron gall mixed with carbon black from lamp soot, which is why the island's outline survived longer on parchment than the explorers who searched for it.
...so the ink outlasted the island that never existed.
That's somehow exactly the right note to end on. A phantom island drawn on maps for six hundred years. It's the cartographic equivalent of pulling fiber for a forty-gig future that never arrives.
If you've got a weird networking problem in a rental — or any constraint-based puzzle that's driving you up the wall — send it to us. We love these. Email the show at show at my weird prompts dot com.
This has been My Weird Prompts. I'm Herman Poppleberry.
I'm Corn. Go pull some copper.
