Why Things Take Longer Than You Think

I'm going to have to write a longer piece on this, but I wanted to put out some key pointers before I do.  This has developed in my head from what has become another _frank and open exchange of views_ that I'm having with

lprovenover on his LJ.  

Full disclosure: I work in the mobile phone business.  My company are a provider of engineering services to OEMs (the people that make phones as their own brand; Nokia, Samsung, etc...), ODMs (the mostly Asian box shifters who throw stuff together and manufacture stuff), Silicon and Radio manufacturers and a bunch of 3rd party software providers.  We used to have our own product line but we recently moved away from that because, well, lots of reasons I won't go into.  My advice, however, is that if you want to make money in the phone business, don't write applications.

The interesting thing about the discussion is that it is almost identical to ones I was having 3-4 years ago with engineering managers from Microsoft about the mobile industry, ok so they were coming from a slightly different angle to Liam but the arguments were pretty much the same.  They were strong PC people, they really understood how OS platforms work, they understand how components come together, so why do the phone people all make it out to be so damn hard.  They were going to explain to them the error of their ways and show them how it was done.  That was 2002 and they were trying to make "Stinger" work, they've released 4 versions of the OS (2 major kernel changes and 2 retreds) and they're working on the 5th which will be a major change and is due in 2008.

They really thought that the problem was the mobile industry and embedded systems engineers working in the phone business didn't really understand a market.  They were imature, they didn't understand hardware integration, if they'd just stop whinging about radio stack timing issues and systems resource problems, then they could settle down and let the real engineers fix it.

There are a couple of problems with this logic.  Firstly, the phone industry has been at this a long time and has shipped more product than the PC industry - pushing 2 billion active handsets around the world now.  Think about that.  30%+ of the world's population has a mobile phone, most of those are GSM.  Secondly, unlike the PC world, most of the phones and the people who use them don't have a PC or access to the internet, so the phone has to just work.

That's the crux of the problem with a mass market consumer good like a phone.  Sure, there's a small number of high end devices which are open or semi open, but the operators don't want people having too much control over that.  Playing, for example, with the QOS settings on a WCDMA phone can tear down a local cell.  Hijacking enough of them and you have a really bad day on your network.  I'm reliably informed that 20,000 trial phones on Cellnet a few years ago pulled down all their Motorola Base Station controllers and that was in the warm cuddly GSM world.  Operators can't afford to let that happen, they have to be assured of a few things before the device attaches to a radio network.

The way this gets tackled at the moment is you either black box the radio - which generally leads to average call performance and phone performance (this is the iPhone approach which I'm told has average voice quality and dropped call rates (not seen yet if that's the phone or AT&T, it'll be interesting to see the European experience) or you tightly integrate the radio and lock the sucker down (virtually every phone in the world).  The reason for this is mostly cost.  Unless you have a device that is soooo cool that people will pay anything (iPhone) then you're worried about bringing the cost down by a lot.  This means reducing the numbers of chips and integrating things tighter.

Building single core application processors and chips that have full global radio modules (GSM, WCDMA, CDMA on 5 bands etc...) is insanely hard.  It's hugely expensive requiring huge man years of engineering effort.

Leading on from this is the other problem with mass market products that get used a lot.  They don't last long.  People change them every 1-2 years.  In the time that I've had this desktop I'm using, I've had 5 mobile phones.  I muck around with the PC all the time, however, I've hardly ever upgraded the firmware on my phones, there's no need and I need it to do one critical thing all the time above all others - make calls.

That's the sad reason why things take so long.  I can see the future, sort of.  I know what the phones look like next Christmas because we're already designing them.  I know what the silicon and radio looks like for the Christmas after that for the same reasons.  I know where the next big evolutions are being built and where they'd penciled into the timelines, and they're going to be in engineering around the turn of the decade.

They'll be some incremental stuff, they'll be some cool things.  IMS services are being rolled out, but they were being tested first FIVE years ago and are now starting to hit the consumer space.  That will drive more work in Open OS phones, but that will lead to more standards and more rigour I suspect.

One thing I am sure of in this day and age.  The governments of the world are not going to allow hackers access to kit that could tear down the mobile communications infrastructure.  But don't worry, that's at least 10 years away ;)