Old LTO flame archives

Flame Hardware Questions
21

true!

22

The “archive to tape” format was pretty simplistic, and definitely pre-dated LFTS:

  • images and audio in “raw” (framestore) format, the equivalent of one giant “.seg” file archive
  • EOF (End Of File) marker
  • the archive metadata “blob” (project/desktop/library/clip metadata)
  • EOF (End Of File) marker
  • EOD (End Of Data) marker

The Achille’s heel of the system was that when you wanted to append to an archive, Flame would have to do the following:

  • seek to the EOF marker
  • read the archive metadata
  • seek back to just before the EOF marker (thus positioning at the end of existing data)
  • write any additional images and audio, overwriting the EOF marker and the previous archive metadata
  • write a new EOF marker
  • rewrite the update archive metadata, then EOF and EOD

(remember, tape is linear, you cannot “insert” anything)

So if the process aborted (Flame crash, system crash, tape problem) during the update, you would be left with an archive tape with no metadata, which might or might not be recoverable using a copy of the archive metadata stored on disk as the Online Table of Content (OTOC, remember that?).

In very old versions of Flame, the File Archive and Tape Archive formats were essentially the same, so you could play tricks with using “dd” to read the audio/video section of the archive into a large “.seg” file, then the archive metadata, and if you named it right, you could fool Flame into thinking that it was a file archive all along, but I think that stopped working at some point.

This illustrates how tricky a problem long term digital archiving is: it’s not just about making sure that the physical media still works, but also having access to the hardware that can read it, and the software that can understand the format of the data on the media. It’s a bit old by now (2007), but the AMPAS white paper The Digital Dilemma talks about these issues at length.

2 Likes
23

The volatility during the reading or writing process is why I never appended an LTO unless I still had all the data on the flame and could start over if need be. Also the problem goes beyond just being able to recover from the physical media. We still need software that can make use of the data we recover. I can load flame archives that are a dozen years old, but many of the features I used then are no longer supported.