As a co-founder of ITA Software (travel search) and founder of the company that makes <a href="http://inky.com" rel="nofollow">http://inky.com</a> (email), I often see unexpected parallels between travel search and email. In both spaces we get periodic design documents and slide ware that look really cool and get everyone all excited. But in both spaces these designs rarely get implemented, much less reach production.<p>The reason is that the domain details are so extensive and difficult that almost nobody can get past them. I like to describe both these problem domains as "fractal," because when you're 100,000 feet up it looks pretty simple, but the closer you get the more details there are. If I had a dollar for every hacker who told me how easy it would be to make a travel search web site or an email client, I'd be rich.<p>In both spaces, a very small number of players create the core technologies, and a much larger set of players layer on top of these cores. In travel, for example, you have the GDS companies (Sabre, Amadeus, etc.), ITA, and Expedia. Everybody else -- literally, <i>everybody</i> else -- layers on top of one these systems. Kayak? ITA customer. Hipmunk? ITA customer. Etc. This means the vast majority of players don't actually have to deal with the fractal domain complexity. And almost nobody has any clue there's any difference between say, an ITA and a Hipmunk, even though one has a million lines of code and runs thousands of machines and the other has a fairly standard website. (#)<p>Similarly, in email, a handful of players create real mail stacks. These are the usual suspects (Microsoft, Google, Apple, IBM) and a handful of others (Inky, Sparrow, Thunderbird) Everybody else -- literally <i>everybody</i> else -- layers on top of GMail or Outlook. Mailbox.app? Layer on top of GMail. Xobni? Layer on top of Outlook. Postbox? Fork of Thunderbird.<p>In the mail space, there are at least some decent open source libraries to use. In the travel search space there is literally <i>nothing</i> to start from but a blank sheet of paper. Carl de Marcken did most of the early work figuring it out for ITA, and it nearly killed him.<p>I actually seek out domains like this; either because they're defensible, or because I'm insane. I'm not sure which yet.<p>(#) If you're thinking it's necessary or even better to own the million lines of code and run thousands of machines, think again: Kayak created substantially more value for its shareholders than ITA did, with a lot less effort.