Before I move on to other topics for a while (next week is Siggraph, the coolest trade show of the year as far as I'm concerned), I want to describe some of the ways in which the Apple Newton fell short. I'll also explain how these deficiencies relate to today's similar devices-- PDAs, smartphones, and tablets.
As with my post yesterday, these comments are drawn from notes I made during the seven years I used a Newton MessagePad 2100.
Very early on in my Newton experience, I made a simple comment: "Yes, it's too large." The MP2100 was huge, nearly the size of a VHS videotape cartridge. (I wonder if it isn't already too late to be using that reference...) It was much too large for any ordinary pocket and even for most fanny packs. A big overcoat pocket would work. Normally, if I wasn't taking my briefcase somewhere, I didn't take the Newton. The Treo 650 I carry today is more effective for short notes because I'm basically never without it.
On the other hand, I don't think I'd find much use for a PDA-like device much smaller than the Treo. I think the Treo is just about the perfect size for this kind of device (the iPhone, probably not by coincidence, is almost exactly the same size).
As good as the recognizer was for whole words, it was very poor on single letters. The context of a whole word really helped the recognizer figure out what each letter was meant to be. For example, a capital "H" was frequently mis-recognized as "It" or "tl". A lower-case "s" almost always turned into a capital "S" instead. Symbol characters were even worse. The "%" and "/" symbols were virtually impossible for me to enter.
Fortunately, the Newton recognizer could be programmed to replace one string with another. Although meant for purposes such as replacing "sinc" with "Sincerely Yours," it could also be used to override these inappropriate recognizer behaviors. I had "o/o", "°/0", and "°/O" programmed to result in "%". This feature didn't solve the problems but caught the most common problems.
Even these minor problems sometimes took the fun out of handwriting recognition. I could write a page of straight text with few if any errors, but trying to write down anything more complicated was a pain.
My Tablet PC has a similar problem even today, and indeed, this may not be a solved problem in any current handwriting recognizer. URLs are commonly transcribed by hand, and many people have to write HTML (I have to type HTML code into this blog, for example). The difficulty of recognizing isolated symbols may be a small factor in discouraging wider use of handwriting recognition.
At first, the MP2100 had only the standard serial port for input and output. The port supported AppleTalk, a 240-kbps networking mode, but that was very slow for synchronizing the Newton with my desktop Mac. Eventually Apple added software support for Ethernet PC Cards, which helped a lot, but what this device really needed was USB. It was just a few years too early for that.
Today, the comparable requirement is for WiFi, although I think the new Wireless USB standard may become necessary as well. Unfortunately, neither of these provides the secondary function of USB-- that of providing power.
USB doesn't provide enough power to run a hard disk or fast-charge a large lithium battery, so I think we really need a new wired interface-- maybe Gigabit Ethernet with a 12V, 1A power supply? Something for device designers to think about, anyway.
Apple offered a separate keyboard for users who wanted to do a lot of typing on the Newton. It was basically a laptop-style keyboard, but with fewer keys than most. Still it was much bigger than the Newton, and couldn't fold or roll up to save space. Unlike a PC or Mac keyboard, the Newton keyboard output straight serial ASCII characters, so you couldn't just borrow a keyboard at your destination; if you needed one, you had to bring it with you. (Oddly, Newton keyboards briefly found another life during the heyday of the Palm PDA, which also supported a serial ASCII keyboard.)
If I was going to turn the iPhone into a Newton-like PDA, even with greatly improved handwriting recognition, I'd still want to have a small keyboard available... either built in (fold-out, slide-out, whatever) or separate but attachable.
The MP2100 had a folding cover over the screen, but somehow Apple missed the notion of using the cover as a stand to prop the unit up. It almost worked anyway. Even today, hardly any PDAs, smartphones, or slate-style tablet PCs will support themselves on a tabletop. It seems like a curious omission, since this is so easy to do.
Apple gave the MP2100 the ability to record and play back audio, but there was never any MP3 support, the speaker was weak, and Apple never supported the audio in/out connections it designed into the serial-port connector. Even just a headset-jack dongle would have been really useful.
Lesson: if you go to the trouble of designing in a hardware feature, do it right, and support it with software.
Generally, Apple made it really difficult for third parties to support the Newton with hardware add-ons. The company really learned that lesson with the iPod, though...
Synchronization with a desktop Mac or PC was always a pain, and just got worse and worse during the lifespan of the MP2100.
In fact, I don't know of a PDA or smartphone that does a really great job of synchronization. Some are good enough, but that's about it.
The MP2100 would have made a fine eBook reader, but Apple never gave that application more than cursory support. It's such a useful feature that it ought to be regarded as absolutely necessary. And once that feature is present, every machine ought to come with a complete, easy-to-read user manual for itself, every available accessory, and as many compatible peripherals as possible. All this documentation may require another few dollars worth of flash memory to hold it, but it would probably pay for itself by encouraging accessory purchases.
Any time I brought out the Newton in company, there was a pretty good chance that someone would ask to try the handwriting recognition. There was a Guest mode in the recognizer, but what the machine really needed was a Demo mode. The iPhone could use this today-- a mode in which all the real user data would be hidden, making the system look like a brand-new machine. Then, when exiting Demo mode, everything the guest did can be erased.
At one point, I broke the surface glass on my MP2100's LCD. The LCD itself was cracked but most of the screen was still visible. I wanted to do a backup before sending the unit off to be repaired, but the broken glass under the digitizer film kept that from working properly, the Newton keyboard couldn't do much, and there was never any such thing as a Newton mouse. I was able to perform the backup by swapping logic boards with a borrowed Newton, but that's not going to be an option for most people.
It seems to me that complex devices need simple backup methods for these situations. One solution that occurred to me at the time would be a PC Card (these days, an SD card or USB thumb drive would be a better solution) that is pre-authorized to perform backups. Just insert the card and boot the machine; as long as the machine can boot far enough to see the card, it would then proceed to back up as much recent data as possible.
Well, that's enough about the Newton for now.
I strongly, strongly recommend that anyone developing a handheld device with PDA-like functionality buy a MessagePad 2100 (they're readily available on eBay, usually for $150 or less) and study it. Use it enough that you come to understand why it works the way it does, and what's both good and bad about it. It distresses me to see new devices being introduced that don't have some of the good features from the Newton... and don't solve some of the Newton's problems. That's just a shame.