The brief nobody had written before
The technical delivery was contracted to Ashlawn Data Systems Ltd as prime contractor, who outsourced the engineering to OpusVL. It covered the entire digital infrastructure: the online ticket office, the call centre operations, the document scanning and indexing pipeline, and the archive itself. The whole platform was built from scratch in twelve months. There was no Shopify, no Stripe, no AWS; in modern terms it was a cloud-based real-time e-commerce platform, years before anyone used the word "cloud".
But the most consequential part of the brief was quieter: the data collected would need to be readable in 2101.
The hundred-year format problem
Consider what a century means for digital information. A hundred years before the project launched, Queen Victoria had just died and the Wright brothers had not yet flown. The format decisions made in 2001 would need to survive a comparable span of change.
The question the team asked was simple, with profound implications: if someone opens this archive in 2101, what do they need in order to understand what is inside it? The answer was not just the data. It was the ability to read the code that structured the data.
The argument: proprietary or open?
In 2001, open source was not the default choice for a project like this. The commercial world ran on proprietary software, and it was argued at the time that the Odyssey system should be built on a Microsoft platform - the safe, conventional choice of the day.
The counter-argument, put by the technical lead, was decisive: nobody could assure that the software would still be available in 2101. If the software that created and structured the archive were proprietary, the reader in 2101 would need access to that specific software - or would have to reverse-engineer the format. Neither could be guaranteed over a century. A vendor's continued existence, let alone its goodwill, cannot be promised a hundred years out.
Open source resolved the problem completely. The source code itself becomes the documentation: anyone holding the code can read it, understand the data structures, and extract the content - no vendor required. This was not an ideological position; it was an engineering decision about long-term readability, designed for an audience nobody alive would ever meet.
The choices
- Stack
- Linux operating system, Apache web server, PostgreSQL database, Perl application code - all open source, and all still actively maintained decades after they were chosen.
- Formats
- TIFF for digitised images - uncompressed, thoroughly documented, and already stable for over a decade in 2001. Plain text and XML for structured data. No proprietary database dumps, no application-specific file formats - everything readable with basic tools.
- Principle
- Data should be self-describing and readable without the creating software. In 2001 this was simply common sense applied to an uncommon timescale; the digital preservation community has since formalised it. The OAIS reference model (ISO 14721, published 2002 - a year after Odyssey launched) established that preserved information must be independently understandable. The Smithsonian recommends formats that are "open, standard, non-proprietary, and well-established". Software Heritage, the UNESCO-backed initiative, preserves source code precisely because it contains human-readable knowledge. Odyssey worked from first principles; the field caught up.
The operational system
Much of what the platform did had never been done before: real-time ticket production, a call centre and website operating as one online system, payments authorised in real time. It was a groundbreaking project, assembled from parts that did not yet exist.
The platform sold tickets through three channels - online, telephone and post - because its passengers were veterans in their seventies, eighties and nineties, many of whom had never used a computer. Accessibility was a design foundation, not an afterthought: large-character interfaces, a prominent text-size control on every page (this site keeps one in its footer, in homage), an SOS help button on every screen, and parallel paper channels managed by the same system.
When a ticket was purchased - on the website, or by a call centre operator at Richmond taking a telephone order - the payment was authorised in real time. There were no payment service providers in 2001, and no payment gateway APIs: the system interfaced directly with the bank, over a modem running at 300 baud. The speed was a deliberate engineering choice - a 300-baud connection spends no time negotiating, so an authorisation completed faster end to end than it would have at nominally higher speeds. The bridge to the bank's systems was built by hand.
On screen, the website showed the passenger their own ticket, live, as it was created - rendered in the browser with their details in place, using web techniques that were not yet in common use in 2001. What they saw was what the printers at Richmond produced.
The authorised sale then drove the printers at the Home in real time, through printer drivers written from scratch, printing the unique ticket number onto the ticket and the purchaser's address onto a label. The tickets were pre-printed with the graphical design and overprinted with the number at the moment of sale - which made the design of the ticket itself special: it had to leave a non-laminated space where the overprint would take. That, too, was deliberate. Beyond the pre-printed stock and labels, the Administration Office ran with zero consumables, giving the operation the best possible chance of continuous running. The operators' job was reduced to the minimum - insert the ticket, label the envelope, post it with any enclosures - and the finished ticket, with its unique number and PIN, was posted the same day.
Inbound submissions - the stories, documents and photographs arriving by post - were digitised on arrival through a bespoke scanner interface with structured indexing. Some of that material was sensitive, including information still subject to the Official Secrets Act, and security of the data was paramount throughout. Separate interfaces served operators, administrators and the participants themselves, who had a full year to edit, extend or withdraw their submissions before the archive was sealed.
The Home operated everything remotely over a leased line installed specifically for the project, with the system hosted centrally - genuine real-time remote peripheral interfacing, in 2001, over a connection specified and commissioned for the purpose.
Aged well
The modem-based payment processing is now a museum piece; the leased line has been replaced by ubiquitous broadband; building an e-commerce system from scratch seems quaint. But the archive format decisions have aged perfectly:
- TIFF remains a preservation standard recommended by the Smithsonian and the Library of Congress.
- XML remains the foundation of structured data interchange.
- PostgreSQL is one of the most widely used databases in the world.
- Perl remains readable and actively maintained.
- Linux runs most of the world's servers.
The principle that guided the choices - data stored in formats anyone can read, using software anyone can inspect - is now considered best practice by every major preservation institution. The original Perl source code for the Odyssey Timeship system still exists, kept with the project's records - code written so that a hundred-year archive could be read, preserved so that its readers can find it.
What happens at journey's end is on the archive and 2101.