JesseW on IRC, and various other wikis. Archivist by passion, software dev by profession.
Six categories of backups
All of which there can be multiple instances of. Also, the copies in various places may not be identical, as long as they are in a general sense "the same object".
- Online, in a well-known location
- Somewhere anyone can get a copy immediately, and nearly anyone who cares about the subject would know how to find it. Great for accessibility, and reliability (if it ceases to be available, people will know quickly and definitively), but most vulnerable to attacks by people who object to the material.
- Online, in an obscure location
- Anyone who knows where it is can get a copy, but only a few people, generally those deeply involved with the topic, know where to look. Serves as a backup for the well-known location(s), which allows an unbounded group of people to re-stock the well-known locations, without explicitly having to communicate between themselves. Loses some accessibility and reliability (since fewer people will be checking that it is still available).
- Online, but lost
- Even the custodian hosting it isn't aware of what it is -- but anyone is still free to get a copy of it, should they come across it, or decide to copy things indiscriminately. A strange case, pretty close to useless for accessibility, but useful to defend against targeted attempts to destroy copies -- if no human is aware that a copy is located in a particular place, no human can be pressured into removing it. Still vulnerable to automated searches -- if it is an bit-for-bit duplicate, and an automated search is made for such duplicates, it can be found (that's what "online" means). Quite vulnerable to accidental loss, as keeping things online costs some money (although often not much).
- Offline, but the custodian is well known, and well known to have it
- An example of this would be, "Everybody knows Joe has a copy of it. If you ask him for a copy, he might or might not give it to you, and you have to ask first -- but you can be pretty confident the copy exists." Not bad for accessibility, and much less vulnerable to automated attacks against copies; but still equally vulnerable to target attacks against the custodian. Higher risk of accidental loss, as only the custodian can verify accessibility.
- Offline, and the custodian (and/or their possession of a copy) is obscure
- Example: "I heard someone who was known as John D might have a copy, if you can find him to ask." Not great for accessibility, but resistant to both automated and casual attacks on the custodian. Still vulnerable to determined investigators -- the custodian is known to some others, and the fact of whether or not they have a copy is known to themselves, so they could be pressured to destroy it, or cease distributing it. May have slightly less risk of accidental loss as compared with a well-known custodian, as the custodian may be more passionate about the topic. Still high risk of accidental loss, due to not being online.
- Offline, and lost
- Examples: On a hard-drive, or CD, or floppy, or print-out, stuck in a drawer or an attic. Has the same great strength against intentional attacks as the online-but-lost case, with the additional benefit of being invulnerable to automated searches, and having lower continuing costs (no need to pay for server electricity or data connection). However, such copies are pretty much inaccessible, and quite vulnerable to accidental loss whenever they are noticed (as their contents (and value) are unknown to their custodians).
The more of these categories that contain copies of a given object, the better for its preservation.