Two strategies
Migration converts a file into a newer format that current software can open. Emulation leaves the file exactly as it was and recreates, in software, the machine that once read it.
Migration is simpler for the person using the file. Emulation is truer to the original. Lazarus does both, and always keeps the original either way, so that a decision made today can be revisited when better tools exist.
When Lazarus migrates
Material whose content can be separated from the software that made it is migrated: documents, photographs, sound, video and data.
| Material | Migrated to | Kept alongside |
|---|---|---|
| Word-processor documents | PDF/A-2b | The original files |
| EML messages and PDF/A-2b | The original mailboxes | |
| Photographs and scans | TIFF | The originals as taken |
| Audio from DAT and MiniDisc | FLAC, 24-bit | The raw transfer |
| Digital video tape | FFV1 video in Matroska | Logs of every dropout |
| Instrument and survey data | CSV with a data dictionary | The original files |
| Websites since 2004 | WARC, replayed in the reading room | The server images |
Migration always loses something. Fonts are substituted, layouts shift, a spreadsheet’s formulas become fixed numbers. The original is kept so that the loss is never permanent.
When Lazarus emulates
Software is behavior, and behavior cannot be migrated. Games, demos, applications, interactive works and the early web are kept in their original form and run in emulation, through the emulation toolkit.
Each platform has an emulation profile: an emulator core, the system software and firmware the machine needed, its configuration and a set of tests the profile must pass. Every record that runs in emulation names the profile it uses.
What has to survive
Archivists call the qualities a work cannot lose without becoming something else its significant properties. They differ by kind of work. For a game, timing, sound and input matter most; for a demo, exact timing down to the cycle; for a spreadsheet, the formulas rather than the numbers. Profiles are validated against these properties:
- Speed is matched to the original hardware within a tenth of a percent.
- Sound is compared with recordings made on real machines.
- Picture is shown as the machine produced it, with its original aspect ratio, and optionally through a filter that approximates the screen it was made for.
- Input delay is measured, and kept close to the original.
What emulation cannot do
Some things do not survive in either form. Peripherals such as light guns and pen plotters can be imitated but not replaced. Online games whose servers have shut down can be preserved only in part. And no emulator can reproduce the weight of a keyboard or the sound of a drive seeking. Where that matters, Lazarus records it, in photographs, video and sound, and keeps the recording with the record.