Hacker News Reader

Comment by joshuaheard | Hacker News Reader

This and all the stories in the news show that the airline industry needs some serious updating.

For whatever reason, mainframe systems are so deeply entrenched, that all of the major Global Distribution System [0] providers seem to have their legacy systems wrapped in modern "serialization facades" to produce the XML or JSON that gets consumed by external systems.

There are probably a hundred ways to rationalize why. Inertia, lock-in, obstinate grey breards refusing to retire and holding infrastructure hostage until they die... , if-it-ain't-broke-don't-fix-it-even-when-it-is-broke, "we can't afford to stop the world for your cargo-cult modernization rewrite, sonny...", too big to fail, on and on...

It's a pretty weird situation, but by my estimate, it's probably going to require totally new airlines rolling their own systems from scratch, in parallel isolation, next to the huge legacy carriers as they continue to operate, in order to transition, modernize and unseat the current technology. That seems to be the way things work in industries this big.

Fast food, banking, telecommunications. All of them have examples of modern upstarts that are now world-wide staples, simply because they started from scratch with more modern efficient systems, as they continue to co-exist alongside dinosaurs that still haven't gone away.

[0] https://en.wikipedia.org/wiki/Global_distribution_system

>There are probably a hundred ways to rationalize why

It's just a difficult problem to solve. Many capable organizations have tried in earnest, including one Google bought for $700M. Thus far, the non TPF solutions can only scale to medium sized airlines.

It's not impossible, it's just a huge investment, and the payoff is lower than other potential uses of the same money. That's why all previous efforts stalled.

FWIW, Amadeus has been working for years and years on a migration away from IBM mainframes. The project took hundreds of man-years and it's nearing completion very soon (I mean, maybe this or next year, that's "very soon" in the GDS scale).

But as you can imagine, migrating legacy systems to new hardware with the spec "it should work as the old system" is not the most sexy job and not the easiest to recruit great programmers for.

(I work for Amadeus, but on unrelated projects - though I have friends who worked on the migration).

You can read more here for example: https://www.nextplatform.com/2015/08/04/amadeus-takes-off-wi...

> it's nearing completion very soon (I mean, maybe this or next year, that's "very soon" in the GDS scale).

On whose scale is this or next year not very soon?

It feels like it's a classic case of where coming first can hurt you down the road.

For example, part of the reason the US is so far behind on credit card technology (just getting around to adopting EMV) is because the US was the first country to widely adopt credit cards in the first place and there's a lot of inertia.

In the airlines case, they were one of the first industries to computerize. Now they're stuck with the mainframes. Same thing happened in banking.

The credit card point doesn't make sense. The UK and Australia have very high credit card use, but both have had EMV for well over a decade -- I think the UK was first with the modern standard.

The argument makes more sense for infrastructure that needs upgrading, like Internet access, railways, sewers, water pipes.

That's silly. There's huge inertia in the US because they're all insanely profitable and there's no competition.

There is no reason the SSR codes shouldn't work as a solution. This is less about underlying tech, and more about specific effort to use it correctly to solve the issues.

Reply on news.ycombinator.com