What is actually worth putting in an offline library
Most offline archives fail the same way: too much material, too little of it useful. Here is how to decide what earns space in yours — whether you are filling a USB stick you already own or buying a finished case.
The arithmetic comes first
Storage stopped being the problem years ago, so stop spending it like it is. A plain-text book is roughly half a megabyte. The same 300-page manual photographed page by page runs 100–300 MB — a few hundred times the space, and you cannot search it with a phrase. Put the formats side by side and the choice is not close.
| Format | One 300-page manual | Search a phrase inside it | Still opens in twenty years |
|---|---|---|---|
| Plain UTF-8 text | About 0.6 MB | Yes, in about a second | Yes — anything with a text editor |
| EPUB or office document | 1–5 MB | Yes, if the reader has search | Usually — needs the reader app |
| PDF with a text layer | 20–60 MB | Yes, with character errors from scanning | Yes, if a PDF reader exists |
| Scanned PDF, images only | 100–300 MB | No — you scroll and squint | Yes, and still unsearchable |
| Video, one hour | 500 MB–2 GB | No | Only if the codec still exists |
| Audio, one hour | 50–100 MB | Only if someone typed it out | Usually |
Now the sum that matters: twenty-nine gigabytes of plain text is on the order of forty thousand books. Nobody reads forty thousand books, and nobody has forty thousand problems. A well-chosen 29 GB beats several hundred gigabytes of scans and lecture video, because it is small enough to search and carry forward. Selection is the scarce thing, not storage.
Plain text and searchable formats beat pictures
Four properties matter more than fidelity.
- Search. You do not want a manual, you want the one paragraph about your well pump. An index over text answers in a second; a shelf of scanned files means scrolling by eye.
- Portability. Plain text opens on a 2011 laptop with no network, an e-reader, or a phone — no login, no licence check, no account. Apps that need updating eventually stop.
- Damage. A truncated text file still opens and you lose the last few pages. A truncated video file is often nothing at all.
- Longevity. UTF-8 text is an openly documented convention with hundreds of implementations. Video leans on software that is far larger and far younger.
Keep a few images where the picture is the point: wiring diagrams, welding position charts, plant identification, map sheets. Everything made of words should be text.
What earns the space
Judge a category by whether it answers a question you will have within two weeks of losing power, the road, or a technician. These ten hold up.
| Category | What to look for |
|---|---|
| Field manuals and repair guides | Service manuals for the machines you own — generator, tractor, chainsaw, pump. Model-specific beats general. |
| Medical references and first aid | Current editions with a working index, kept to read alongside professional care, never instead of it. |
| Water and sanitation | Pump and well repair, cistern storage, latrine and greywater handling, simple testing. |
| Food preservation and canning | One authoritative preserving guide plus the extension literature for your region. Pressure canning procedure, not blog recipes. |
| Welding, machining, electrical | Process and filler charts, feeds and speeds, motor and panel references, code basics for Ontario. |
| Veterinary and livestock | Species-specific husbandry, housing, nutrition, and how to describe a problem to a vet by phone. |
| Maps | Topographic coverage of your own area at useful scale, plus a road atlas. Map data is the biggest honest consumer of space. |
| Legal and government forms | Insurance, land and permit paperwork, with fill-in copies you can print. |
| Seed saving | Isolation distances, pollination habits, storage and viability for the varieties you actually grow. |
| Radio | Band plans, antenna arithmetic, licence material, operating practice for the bands you can reach. |
What does not earn the space
- General encyclopedias. Enormous, shallow, and rarely the fastest route to something you can act on.
- News archives. Thousands of pages about events you cannot change, with no instructions in them.
- Video lectures and course bundles. Linear and unsearchable; you cannot skim an hour of video for one torque figure.
- Forum dumps. Ten unverified opinions per question and no way to tell which poster has done the job.
- Duplicates. Three editions of one manual mean three answers and no way to know which is current. Keep one, note its printing date.
- Anything you will never open. If it has not been opened in a year of normal life, it will not be opened in the dark.
One good edition beats four mediocre ones
The instinct when storage is cheap is to take every copy. Resist it: on a specific job, conflicting instructions are worse than none, because you cannot tell which page is wrong while you are holding a wrench. Choose by these four things, in this order.
- Authority. A named agency, institution or long-standing publisher beats an unattributed file. Extension services and standards bodies beat photocopied handouts.
- Date, where it matters. For wiring and pressure canning, take the newest edition. For hand tools, forging or timber framing, a 1940s text often beats a modern one stripped of detail.
- An index and real headings. Open it and check. If the contents page is a photograph, so is everything else, and search will fail.
- Words, not pictures of words. Search a phrase you know is in the book. If nothing comes back, you have an image collection.
Sixty-second check on a candidate file
Search one distinctive phrase and confirm it lands on the right page. Read a paragraph at random — a scan that renders litres as “1itres” will hide the answer from you. Check the printing date. Then delete the older copy you were keeping just in case.
Test your archive with the cord pulled
An unverified archive is a hope. This takes twenty minutes.
- Search a phrase, not a word — something that appears once. Confirm it lands on that page.
- Turn off Wi-Fi, reboot, search again. Anything needing a login, a licence check or a download is not offline.
- Open a book on a phone with a USB adapter. If it needs a laptop, you have a laptop library.
- Open three files at random. If the names are strings of digits, you will not find anything by name either.
- Unplug the drive mid-session, plug it back in, search again. Cheap media fails; know how.
- Print the ten pages you would want at three in the morning, and put them where you would look. Paper has never needed a battery.
Last, test the part everyone skips: ask a question and read the answer. A reply that reads like an instruction but names no source is a guess wearing a uniform. An answer that says “pressure canning, Ball Blue Book, 37th edition, page 62” lets you open page 62, check the numbers yourself, and see whether the guidance fits your altitude and your jars. That is the whole difference: one you can verify in ninety seconds, the other you can only trust or doubt. A library that names its source is doing what a library is for — handing you the page, not an opinion.
On medical references: keep current first aid and clinical material in the archive, and treat it as reading to consult alongside professional care, never as a substitute for it. Print your local emergency numbers and tape them inside the lid.
If you are building your own archive, the winter grid-down playbook and the family preparedness audit both assume one, and the measure-your-enclosure guide covers the case side.
COLDSTART ANCHOR is the other route: a sealed aluminium case built to order in Cambridge, Ontario, with about 29 GB of plain text inside — full-text search, offline maps, field manuals and medical references, plus a small local assistant that answers from the books and cites the book and page. No internet, no account, no expiry date. Batch 01 is 100 units; the Core Kit is $519 and the Shielded Case is $1,819, and the price shown is the total payable. Terms are on the Batch 01 page and in the terms.