The truth about the volcano and the Internet

Don't confuse headlines with truth or bad design and poor customer service with actual failure...

I came across a blog suggesting that the Internet had failed stranded travellers and having been a stranded traveller, I disagree. Bad travel sites made by companies that don't understand what their business is failed stranded travellers. The mainstream news failed stranded travellers (the BBC gave us the distinct impression Calais would be like a refugee camp - it was more like a wet weekend in Wales). The Internet, however, told us what routes to try, got us email confirmation of hotel bookings and navigated us across Spain and France - and let us natter with friends as we went. 

As regular readers know, we just brought ourselves back from Barcelona and I say you should stop blaming 'the Internet' and blame badly designed backends, poorly prepared travel companies with lousy customer service and news sites that dramatise the situation. If you get to Calais you don't need an amateur flotilla, you can just buy a walk-on ticket to the ferry (this is easier at 4am in the morning but actually we saw more people there in the dark than at 1pm for the 2pm sailing ;we went back to the hotel to sleep but we could have taken the 5.30am sailing). I had been making jokes about getting home from Dunkirk in fishing boats: we think we actually saw the flotilla boys (one of whom is a son of a newsreader who's presumably been telling the world it's a catastrophe in Calais); they asked us if we had a ticket home and when we said yes they replied 'shame'. Having driven for 13 hours and slept for about 6, that comment left us with a snarky feeling of Schadenfreude that they only picked up 25 people in their little boats.

There have been far too many people telling people things they don't personally know (the ports are closed, there are no extra trains and other scare stories). There have been some terribly designed Web sites with some backends that can't cope with the amount of inquiries; Eurostar's back end couldn't update the reservations system fast enough to keep up with sales and if P&O's Web site hadn't said no foot passengers would be allowed until Monday the journey would have been less stressful. Leaving the booking to third parties only works when they're good - Amex Travel did a great job for us finding rental cars and hotels and telling us which booking systems were broken, which were working and that yes, it was worth going to Calais. There's been far too much commentary about how the sky is falling and far too little practical information but that's the fault of governments, scaremongering media and clueless travel companies - and how they use the Internet - but not the Internet .

And by the way, the Web is not the Internet; the Web is just a shiny front end written in one of the least effective programming languages I know of to back-end systems that can be anything from a grownup database to a bag of wet string. We happen to know that one of the key booking sites for people trying to get home used to involve three incompatible versions of a Big Name Database (now fixed but I hate to think how the old site would have melted down). As we've been talking to browser developers and Web developers about HTML 5 and backwards compatibility between new browsers and old sites , the conviction has been growing on me that actually, the ad hoc, fail fast, anyone-can-do-it Web programming model is broken. Making it easy to do things has made it too hard to do things right.

The principle of small pieces loosely coupled is more sophisticated than it sounds; they have to be self-contained, caching, connection-aware, storage-enabled, identity-friendly pieces with journaling and transactions. Web sites have to have state - to know who you are and what you're in the middle of dong - and faking it with Ajax isn't good enough for real-time systems. It's barely good enough for comments on social networks (if the fact that three out of the four Facebook comments I make in Facebook Lite in IE8 never appears on the page is down to the browser then I think that's proof of the fragility of the model). If we want a disaster-class Internet that doesn't end up being as chancy a method of communicating as sending a text message when you're buried by an earthquake and hoping that a student volunteer translates it correctly, we need real-time protocols and programming models and robust, fault-tolerant systems. If we're going to rely on the Internet, we have to make it easier to make it reliable.

-Mary