Simon’s Blog

Apple’s new MobileMe sync software is giving people some big problems. Big enough problems that they’re quitting using the service. For example, see what Robert Scoble has to say in his post – Warning: your calendar is in danger with Apple’s Mobile Me.   I know a lot of people are disappointed, but the truth is that writing reliable “sync” software suchas MobileMe is a super-hard problem.  It will take a while for Apple to make this reliable. I thought I’d take a minute to explain why it’s such a hard problem.

The root of the complexity is that when you write sync software, you have to write code that interacts with systems that are outside your control.   For each additional system you have to interact with, the difficulty of creating realiable system increases exponentially.     Now, if everyone wrote code with beautifully designed public interfaces, which remained stable, was bug-free, was well-documented, and whose implementations never changed, then there might not be a problem.  However, that’s not the way the world is.    In fact, most software tends to have bugs; mature software products have many versions in production use, all of which have subtle and not-so-subtle differences in behaviour… and when I say “version”, I’m not only talking about things like Outlook 2003 vs Outlook 2007.   Even Outlook 2003 comes in many, many different versions (it’s patched by Windows update very frequently – and each patch creates a different version that behaves differently (sometimes dramatically so).   Often, the systems you’re trying to integrate with aren’t well documented, so you can’t be even sure you’re doing the integration as you’re intending.  And… well… you get the point: serious bugs tend to abound.

This “interaction” problem is why, if you’re running an Enterprise IT function that has to be reliable, you will usually favor putting only one application on each server (these days, a virtualized server).  Why?  Because, there’s actually an interaction problem even with a single application – the interaction between the OS and the application.   The application might require particular patches to be applied to the OS, in order to work reliably.

If you’re a user of software, quite rightly you don’t care about all these problems.  You want the software you use to “just work”.    However, with any new sync product, for all the reasons above, there will inevitably be bugs to start with.   You simply can’t test it effectively ahead of launch because you don’t have enought test cases to work with until it’s in production. So, to all those MobileMe early adopters, I’d say this: please don’t give up on MobileMe.  The MobileMe team need as many examples of different configurations as possible to find fixes for all the problems that occur.