“We have a lot of optimistic engineers – but not enough pessimistic engineers,” reckons David Rosenthal. In the 1980s, Rosenthal designed the NeWS windowing system with James Gosling. In the 1990s he was NVidia’s fourth employee, or really the first person the three co-founders hired. But for the past few years Rosenthal has been tackling an issue that’s very close to Register readers’ hearts – we know from your mail.
On Monday we discussed the permanence of digital media – or more accurately, the lack of it in our story, Digital memories: we can forget them for you wholesale!. Why should any of us trust our family albums to digital media such as a web photo service when we can’t guarantee it’ll be around?
In this age of wiki-fiddlers and other careless coders, where the line between marketing hypester and developer has been blurred, the premium that’s placed on data integrity seems to be at an all time low. And the lossiest seem to work in “web services”. Uh, look out!
This is where a glum approach can help, reckons David, who is not at all glum in real life.
“Pessimistic coders tend to be systems engineers who have been thinking about building fault tolerant systems for a long time, systems that are more resistant to human error,” he told us on a UK visit this week. “It’s a bigger problem than just the Web.”
What Rosenthal has been building, with some success, is an open source peer to peer system that preserves digital content, even when powerful agencies want to see it altered or trashed. It’s called LOCKSS – for ‘Lots of Copies Keep Stuff Safe’ – and it’s already in use by 87 libraries. The project home is at Stanford University Library in northern Californian.
It’s only addressing one of the problems of digital media – others include obsolescence and proprietary file formats, for example, but it’s one that few people look at. And it’s a very important problem to fix.
Lock up your datum
Digital formats, particularly web with publishing, are much worse than paper when it comes to preserving content for current and future users, and computers make it much easier to rewrite history, Rosenthal says.
“In 1984 Winston Smith’s job was updating the only copy of history,” he reminds us.
“There’s a whole community of librarians who were desperate for this tool, we discovered,” he says.
It’s a P2P system in which multiple copies are made and the member nodes audit each other, and repair any damage. LOCKSS is both a web cache and a web crawler; the cache never gets flushed and the crawler is permanently scavenging.
“Good quality backup is difficult and expensive to do – and archive quality backup is really expensive and difficult to do”, he says.
Many smaller libraries find getting system administration skills difficult and expensive, so the team turned LOCKSS into a turnkey product that boots off a BSD CD. It’s really a network appliance now, he says.
Initially Rosenthal tried to interest potential users in one of the experimental “giant internet file systems” such as NEC’s Intermemory, now OpenStore.
“We tried to sell them the idea of this big file system where no one knows where the data is – but the librarians didn’t buy this story!”
Anyone can use the source code, but the team wants the system to be used by many more medium sized libraries before embarking on a rewrite of the LOCKSS’ P2P protocols. Did he envisage a time when community groups would be able to use the software?
Rosenthal said he would expect it to be several years before it filtered down to the DIY level.
For now, he’s seen an enthusiastic adoption by the humanities community, and built on early grant sponsorship from the National Science Foundation. LOCKSS is targeted at a specific community of people who take data integrity very seriously indeed. But it also sets a standard to which we must hold these new, highly emergent web services – before they lose all our valuable stuff. Backups aren’t self-organizing.
Find out more about LOCKSS excellent work here.