Slashdot Mirror
Digital Media Archiving Challenges Hollywood
HarryCaul writes "Movies are moving to digital, but what about long-term archiving of the master source materials? Turns out it's harder for digital media than for contemporary analog. Data is being lost, and studios have to learn to cope. Phil Feiner of the AMPAS sci-tech division says when he worked on studio feature films he 'found missing frames or corrupted data on 40% of the data tapes that came in from digital intermediate houses' How to deal with it? Regular migration from old media to new media. Grover Crisp, says Sony has put in a program of migrating every two to three years. Other studios are following suit, but what about indie features? Will we lose films like we lost the originals of the 20s?"
Sounds like a good archival method to me...
Tape backup has a limited shelf life. There is an effect known as "bleed through" that degrades data simply because of the physical layout of magnetic tape backup. This maximum shelf life of data on quality DDS media is about five years, therefore data on tape must be renewed in intervals less than the "best before" date.
Additionally, daily backup tapes (differential or complete) have limited write cycles and must be replaced well before the manufacturers recommended maximum write cycles is exceeded.
Obviously optical media is not an acceptable backup solution, due to its many failure points. Tape is still the best, despite the requirement for periodic renewal.
Digital archiving 101 - get with it people.
There are two problems:
Data loss, where the data is actually lost. This is the equivalent of a scratch on a frame of the master negative. The cure is redundancy.
Obsolescence, where the format becomes difficult to read after a period of time. The cure is lossless copying to new formats over time and/or keeping old equipment around.
Another possible cure the the 2nd problem is to convert it to analog in an "easy to digitize" way.
For example, simply "printing" the movie to 3 black-and-white filmstrips, one for each color, is considered archival. These can be rescanned later if needed. For better archiving, use larger film formats.
Preserve each audio track in an archival analog format as well.
Of course this doesn't preserve all the data that a digital filmmaking process has, but you aren't any worse off than you would have been with an analog film.
If you want to, you can preserve each element of each scene separately, in an analog format or a completely-documented digital format but on an archival media, such as a "paper printout" stored on microfilm. I don't think most movie studios will go to this expense.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
This isn't news -- at least not to those of us who deal with data.
The typical procedure is to do a media refresh (ie, copy it) every few years, and to check for damage. There are concepts like LOCKSS (Lots of Copies Keeps Stuff Safe), so those joking about BitTorrent aren't that far off, but it's a little more structured than that.
Dan Cohen gave a talk recently on "Can Today's Scientific Data Be Preserved? The Specter of a 'Digital Dark Age'", which touched on not only the issue of media failure, but also the loss of the knowledge to extract the encoded information. (much like the 'lost languages' that we don't understand now, how do we make sure that future generations have the necessary hardware and software to get the data back out?)
What's disapointing is just how fast the media is failing. Vendors give a 'mean time to failure' estimate that's based on perfect storage, and that they have no real ways of testing (because, well, if you say it's 40 years, are we going to have to wait 40 years before using it?). Even if you're duplicating your tapes, what happens when all of the copies were put on the same potentially bad batch of tapes?
Quite likely, we're going to lose data. And some of it's going to be because we no longer have copies of the data. The rest is going to be lost because there's so much crap being saved that doesn't need to be that we can't find stuff that still has value in the future.
Build it, and they will come^Hplain.
Feel free to disagree but don't tell us, tell the BBC who've relied on the assistance of private collectors to reconstruct their archives.
At the risk of being redundant, the idea of distributed storage is good. Tie in the concepts of ubiquitous computing, prolific high speed internet, vast amounts of unused HD space and clock cycles - Why not?
Doesn't google use massively distributed and redundant storage and processing now?
And when I say why not, I'm not talking about putting it in a rootkit, I'm talking about something along the lines of seti@home + bit torrent + whatever = a global data ether of our collective digital valuables forever being stored by everyone.
Flappinbooger isn't my real name
Just back from Tinsel Town after talking to some of the dudes in the article. Still jetlagged, so it feels a bit unreal reading about it in Slashdot. Still, I'll do my best to explain why things are the way they are...
A film is not often made by a single body. If you are shooting to film, then this will get handled by an editorial department. You may have a fast telecine scan for reviewing the material as dailies. Some of these scans may be used as low-resolution proxies for initial grades. Some chosen bits of film may get re-scanned on a slower pin-resolution scanner for inclusion in the final film. Artificial rendered scenes and special effects may be done by specialist houses, then composited in a post-production house. Your film may have 25 4K images per second in the final version, but the data used to generate it is scattered over the place - if you think a good IT department should be backing all this up, then you haven't worked on a film, my friend. As deadlines approach, people may be working stupid hours, and filling up all the available storage. Then the film gets released, and either makes a billion dollars or doesn't. Either way, the tension is off, people take holiday and zonk out. Nobody will be picking over the cutting-room floor or its digital equivalent looking for things that might be useful twenty years from now. By the time people are back from holiday, they don't know or don't care.
Your end product may be big reels of negative film that you send to a film lab to make prints for cinemas. The lab should keep the golden master clean, and make most of the prints from a second copy. This would be a sensible time to make an archival print of the film. The lab can transfer the whole thing to black and white film. Black and white film does not fade, like conventional colour film does, even in the can. You are getting the print lab to do a pretty full backup of the released film when your people have all gone on holiday. These days you need to back up other stuff. The soundtrack is digital. You will have extra data for the releases in different formats (5:4 TV, 16:9 widescreen, IMAX, etcetera). Still, it is a lot better than nothing. But it is not often done.
The other think is to know what to archive. Very little of the newsreel film I had to sit though as a child to get to the cartoon has survived. Key stuff like the Queen's Coronation or the outbreak of WW2 was clearly history, and put on a special shelf, but little of the day to day stuff survives. There is one cache that survived when a cinema closed, and the tins of newsreel went into landfill. The cinema was in Alaska; the landfill was permafrost, and the film was kept in near ideal refrigetrated conditions. Apart from this fluke, it has probably all gone.
There will probably be digital solutions in time. Increasingly, as we have to manage more different sorts of digital data, there is a need to organize and track everything, which ought to mean it is possible to archive all the essential bits that go into any production. Many other people have posted on the problems of knowing what is on (say) a FAT16 Windows 3.1 disk in some 1980's image format. You can keep copying the data to overcome the degradation of the physical medium, but you still have to know what it means. I know of a system for archiving film images, where the people who did the archiving left the company, and one of them took the laptop with them that had the archiving software, so the ability to read the archives went with them. Do you archive the archiving system? Then, do you archive the system that archived that? Yes - basically, that is exactly what people are proposing to do. But it takes a bit of organizing, and we are not there yet.
Film, on the other hand, has visible images. The 35mm format has remained readable for over 100 years. Even where nitrate stock has flowed over time, we still know what shape it ought to have been. Sometimes we can get something back if we want it badly enough.
A simple analogue solution may be to
Break the data into manageable chunks and use SmartPar (or equivalent) to make recovery files that can repair any one of the chunks. The up-side is you don't need many redundant full backups of the entire data set.
The wide-spread adoptage of SmartPar has been significant in enabling the error-free transmission of huge files via USENET News.
http://parchive.sourceforge.net/
"Because this new approach doesn't benefit from like sized files, it drastically extends the potential applications of PAR. Files such as video, music, and other data can remain in a usable format and still have recovery data associated with them.
The technology is based on a 'Reed-Solomon Code' implementation that allows for recovery of any 'X' real data-blocks for 'X' parity data-blocks present. (Data-blocks referring to files OR much smaller virtual slices of files)."
AC