by Jean-Louis Gassée
Simple is hard. Easy is harder. Invisible is hardest. So goes one of the many proverbs of our computer lore. As Apple found out last month with the MobileMe launch misfires, the lofty promise of "Exchange for the rest of us" translated into a user experience that was neither simple nor easy -- in a highly visible way. Four weeks later, the service appears stable but doubts linger: Is Apple able to run a worldwide wireless data synchronization service for tens of millions of users.
What happened and what does it mean for MobileMe's future?
Let's start by decoding the "Launchpad Chicken" phrase. The game of Chicken is one by which two young males test their virility in the following way: from opposite directions, two cars speed towards each other on the same lane of a country road. The one who steers away first obviously lacks cojones and is derisively called chicken. You might ask about brains versus testes but here we are, the chicken is the one who "blinks first". Now, let's turn to the launchpad. Picture the NASA control room before the launch of an expedition to the Moon. Hundreds of (mostly) men in white short-sleeves shirts, pocket protectors and eyeglasses, hunched before screens, keyboards and telephones. Each one monitors a subsystem: left liquid hydrogen tank, backup gyroscopes, main engine telemetry... In the huge air-conditioned control room, five of these men are sweating, something's not quite right with their baby. The temperature keeps rising, the pressure is falling, the telemetry link is weakening. Almost but not quite in the red zone. If the parameters keep drifting like this, they'll have to pick up the red phone. But who wants to be the one who aborts the launch? So, they sweat some more and hope someone else blinks first. There you have it: Launchpad Chicken.
Now, move the imagery to projects with complicated subsystems. You see how the NASA metaphor made its way to Silicon Valley. There is always hope some other engineer will raise a hand and spare me the embarrassment of admitting my part of the project could crash the launch. This is what happened for MobileMe, with a twist on the cojones, so to speak. No one had enough brains and guts to risk humiliation, to raise a hand and say: Chief, we're not ready here, let's stop everything. As a result, MobileMe badly crashed on launch. A couple of weeks later, we have a leak: an "internal" memo from Steve Jobs. The email states the retroactively obvious, the project should have been delayed or at least launched in stages. No less obviously, a new leader is appointed, Eddy Cue, he'll continue to run the iTunes systems as well. Charitably, the deposed MobileMe boss is granted anonymity, he might have been misinformed by his charges, or he might not have asked the right questions at the right times, it doesn't matter anymore.
But, you'll ask, that doesn't tell us what went wrong, which liquid hydrogen tank sprung a leak. This now gets us into two more topics: sync and size. Sync here means keeping information identical, consistent over two or more devices. Less abstractly, for a simple example, I have a phone and a computer, I want their address books to identical or, at least, consistent. On simple cell phones, I use a cable (or a Bluetooth wireless connection) plus software to copy (parts of) my computer address book to the phone. But, wait a minute, I entered numbers on the phone that are not on my computer; I don't want the copy from the computer to wipe out those new numbers. Trouble starts, as if connecting the cell phone to the computer and running the program wasn't buggy enough. Tou want the software to compare the two address books, the phone's and the laptop's and decide what to keep and what to change, on both devices. But what about homonyms, or different numbers for the same person's home? The program, hopefully, raises those "exceptions" and lets a human arbitrate.
We're just warming up. Now picture a more real-life situation. One traveling consultant with one laptop, one smartphone, both carrying mail, address books and calendars and one assistant in the office with a desktop computer. In Microsoft Exchange's lingo, the assistant is a "delegate", has access, including modifications and new entries, to the traveling consultant's data. Everything must be kept identical, consistent, in sync. How is this done?
Using the Exchange server as an example, it keeps the "true" data. And the "clients", meaning the smartphone, the laptop, the assistant's PC submit changes, new mail, an updated appointment, a new contact home phone to the Exchange server. In turn, the server propagates changes to the clients. We say the updates are "pushed" to the smartphone or the laptop, just as they "push" new mail or a new calendar item to the server. You can easily imagine conflict situations: the same appointment changed by the consultant and the assistant, address updates and the like. By now, at least on Exchange, these "exceptions" are well understood and generally well-handled. But it took years of practice. Just as it has taken years for RIM (founded in 1984), the Blackberry (launched in 1999) creators to polish what is the best-selling synchronized smartphone. Details, details and more subtle mistakes and special cases found and fixed. The Blackberry got its stardom from truly delivering the Simple, Easy, Invisible proposition referred to in the beginning of this essay.
MobileMe aspires to deliver a similarly invisible level of synchronization for people who don't have an Exchange server, hence the "Exchange for the rest if us" slogan. But seeing the launch glitches, I wonder how many people at Apple stooped to using a Blackberry with an Exchange account. Doing this would have sobered them a little in advance of the launch, or delayed the whole thing, or tempered the boasts. Shortly after MobileMe's first missteps, Apple publicly and smartly retracted its use of "Push" to describe MobileMe's synchronization and the "Exchange for the rest of us" motto is no longer seen on the company's Web site.
Moving to size: quantity begets nature. At some (often mysterious) point, more of the same becomes something different. One server, ten servers, more of the same. One thousand servers or, in Google's case, running one million servers is of a different nature. Meaning different people with different knowledge and appetites than the ones needed to run a company's email server. If every other iPhone customer wants to sync a PC or Mac with the newly (or old, with the 2.0 software update) purchased iPhone, MobileMe will soon serve millions and, in a not too distant future, tens of millions of iPhones. Besides knowing or not knowing the Buddha of sync, did the MobileMe team have the experience, the knowledge, the appreciation of the "size" problem before them? Very few people in our industry do. Ask Google's rivals why they were trounced by someone coming late to the game but with a better handle on the "size" or "scale" problem. (See this paper from UC Berkeley, where ultra-large scale computing is actively researched, with private industry subsidies.)
In passing, 10 million MobileMe subscriptions at $100/year is a nice piece of change, one billion dollars, worth the trouble.
Let's step back a little. Apple "pushes" somewhere between 100 and 200 megabytes of updates per month to each Mac user. Last week, the iPhone 2.0.1 update was announced, I connected two iPhones within minutes, the 200Mb files were downloaded and installed without a hitch and I haven't heard any blogosphere complaints on the matter. iTunes has sold billions of songs, serves tens of millions of customers everyday and everything works with very few exceptions. In other words, some very large scale Apple systems do work. As discussed above, the iTunes boss (some say slave driver, a meliorative term in context) in now also in charge of MobileMe.
And, last week, parts of the Gmail service were down for 15 hours or so. Last month, Amazon's respected Web Services went down. And, last year, RIM's servers went down for about half a day in the Western Hemisphere, freaking out Wall Street investment bankers and management consultants. Even the best players must endure their share of false notes.
Back to MobileMe today:if you ask subscribers who've never experienced a Blackberry's smooth delivery of sync, they love MobileMe. It works, it's easy to set up and in the simple (most frequent) case of a PC/Mac with an iPhone, it does the wireless (OTA, Off The Air) sync job as now advertised. We'll see how this scales once iPhones are sold in 21 more countries, 43 total starting August 22nd.
Sorry, the document you have requested is not available.