Store Negatives, Drives and Masters for a Move

Pack the originals first, then the drives, then the paperwork. A back catalogue survives a move when every negative sits in its own sleeve inside a labelled folder, every drive has a checksum file and two copies with one off site, and a migration calendar names the year each format gets read and rewritten. Skip any of the three and you find out at the far end, when the client asks for a frame you cannot open. This is the same discipline press archives have used for decades, and the reasoning behind it is documented in the press criticism archive that tracked how newsrooms filed and lost their own material.

What goes in a sleeve, and what goes in a folder

Negatives go in uncoated paper sleeves, emulsion side away from the paper seam, one strip or one sheet per sleeve. Paper breathes and buffers small humidity swings; plastic sleeves trap whatever moisture was in the room on the day you packed. Write on the sleeve in pencil, never ink: ink migrates through paper over years and can print onto the emulsion. The sleeve carries the frame range and the roll number. The folder carries the job, the date and the client. The box carries the shelf range.

Folders are the unit you actually move. A folder holds one job, its sleeves in order, a contact sheet or a scan index, and a one page note saying where the digital files live and what they are called. That note is the difference between a catalogue and a pile. When the boxes arrive at the new place, you can rebuild the shelf order from the folder spines without opening a single sleeve.

The lesson travels: the label is the catalogue, the object is only the object.

How do you checksum a drive without wasting a day?

You checksum once, at the moment the drive is filled, and you never checksum the whole archive again unless a drive is suspected. Use SHA-256, write the manifest to a plain text file next to the data, and keep that file with the drive and in the off site copy. A 4 TB drive takes roughly 40 to 90 minutes to hash on a desktop, depending on the bus; do it overnight and stop thinking about it.

After the first hash, verify on a schedule, not on a whim. Once a year, read the manifest and re hash a sample: 1 percent of files, chosen by the manifest order, not by hand. If the sample is clean, log the date and move on. If one file fails, re hash the whole drive that week. A failed hash is not a dead file yet; it is a file that needs to be restored from the second copy before the second copy is also touched.

Name the manifest after the drive, not after the job: drive-07-2024.sha256. The drive serial goes in the filename too, because serials are what you read off the case when a drive will not mount. Keep a paper list of serials in the box. Paper does not need a driver.

Two copies and one off site: what counts

Two copies means two separate devices, in two separate rooms, on two separate power circuits if you can manage it. A copy on the same desk as the original is not a copy; it is a second chance at the same accident. The off site copy is the one that saves you from fire, flood and theft, and it can be a drive in another building, a drive in a safe deposit box, or a cloud bucket, as long as you can get it back within the time your clients expect.

For a working catalogue, the practical arrangement is: master drive at the bench, working drive on the shelf, off site drive rotated every quarter. Rotate means you physically swap the off site drive with the shelf drive, then re hash the one that came home. A rotation calendar on the wall is enough. Cloud is fine as a third copy, but upload time is real: 1 TB over a 20 Mbit/s line is about 5 days of continuous upload, so start it before you need it.

Do not encrypt the off site copy with a key that lives only on the machine you are moving. Write the key down, seal it, and store it somewhere that is not the box.

What does a migration calendar actually look like?

A migration calendar is a table with four columns: format, last readable date, next action, owner. Format is the container, not the content: 3.5 inch floppy, CD-R, DVD-R, SATA spinning disk, LTO tape, camera raw on a card. Last readable date is the year you last successfully opened a file from that format on hardware you own. Next action is either "verify sample" or "copy to current format and re hash". Owner is a name, not a team.

Set the review interval by the format, not by the calendar year. Spinning disks get a verify every 12 months and a replacement at 5 years of power on time. Optical media gets a verify every 24 months and a copy at 10 years, sooner if the discs were stored warm. Tape gets a verify every 24 months and a copy at 10 years or when the drive is the last one you can service. Camera cards are not archive; they are transfer media, and they get wiped after the second copy is verified.

The calendar only works if the next action has a date. "Copy the 2003 discs" is a wish. "Copy discs 2003-A to 2003-F to drive 09, re hash, log on 14 March" is a task. Put the task in the same place you put client deadlines, because that is the only calendar you actually read.

How do you move the boxes without losing the order?

Photograph the shelf before you touch it. One wide frame per shelf, one close frame per box label. Those frames are your rebuild map, and they cost nothing. Then pack by shelf, not by size: one box per shelf range, the folder order preserved, a copy of the shelf photo taped inside the lid. Number the boxes on two faces with a marker that does not smear.

Drives travel in their own cases, padded, upright, and never in the same box as the paper. A dropped box of paper is annoying; a dropped drive is a restore. Keep the drives with you, not with the removal van, if the distance is short enough to carry them. At the far end, unpack the drives first, mount them, verify the sample, and only then start on the paper. If a drive fails on arrival, you want to know before the boxes are open and the room is full.

What breaks first when a catalogue is left alone

The first thing to go is the naming. Files get renamed by a well meaning assistant, folders get flattened by a copy that used the wrong flag, and the manifest no longer matches the tree. The second is the hardware: the drive that reads the format is the thing that dies, not the format itself. The third is the person: the owner column in the calendar goes stale, and nobody notices until a client asks.

Guard against all three with one habit. Every time you touch the archive, update the manifest, update the calendar, and write the date on the folder. A catalogue that carries its own dates does not need anyone to remember it. That is the whole trick, and it is the same trick the old press desks used when they filed clippings by hand: the date on the folder is the archive.

Sources

• Library of Congress, "Sustainability of Digital Formats": https://www.loc.gov/preservation/digital/formats/
• National Archives (US), "Frequently Asked Questions about Digital Preservation": https://www.archives.gov/preservation/products/faqs
• ISO 18916, "Imaging materials, Processed imaging materials, Photographic activity test for enclosure materials"