Six Software Products in One Year

I bought a Kinesis Advantage keyboard with the intention of reducing my finger pain associated with typing. Obviously, I spend a good portion of every day typing, and my livelihood basically depends on my being able to continue typing.

I also decided to learn the Dvorak layout while I was learning the Kinesis keyboard. I tried learning Dvorak a few years back but gave it up because I was working with a lot of other people’s keyboards as well - it was too inconvenient to keep switching layouts.

On the Kinesis/Dvorak learning process

• Muscle memory is a huge factor in switching to a different keyboard or layout. Even now, when I’m typing on my laptop I instinctively reach for the Enter key with my right thumb, because that’s where it is on the Kinesis.
• The Kinesis makes a lot of bad habits difficult, whether by accident or by design. You can’t really rest your hands on the pads while typing because then you can’t reach all of the keys. You can’t twist your hands around to move them around the layout because the keys are aligned to suit your hands in the home position. When you move your hands around the layout, suddenly they’re un-aligned and awkward.
• I started out using a lot of ‘mental CPU’ time to handle the conversion. In the beginning, it took all of my concentration just to hit the right keys - I had to separate my thinking from my typing.
• While learning Dvorak, I noticed an interesting progression; I started out pressing just single keys at a time. Gradually, I started combining strings of keys into single motions (something I call ‘chording’, which I’ll come back to). This is similar to how a child learns to read - they recognize single letters, expand out to sounds and eventually can string together words.
• I made rapid progress for about three weeks. The first few days were difficult. I only used the Kinesis+Dvorak for a couple of hours each day because it was very frustrating to learn. I’ve had 20 years with a nice solid brain-keyboard link via the keyboard, and suddenly it’s horribly slow and error-prone. After the first few days things settled down a bit and I could manage an entire day’s work on the Kinesis.
• After the first three weeks, progress slowed. I was still improving, but more in the areas of chording and accuracy. Some keys still gave me consistent problems on the Dvorak layout, particularly G and P.
• You need to sit a little higher than with a normal keyboard. This is a problem for me - I already find standard office chairs too short (even the Aeron!). When time allows, I’ll be buying a nicer chair and getting the gas-lift swapped for a taller one.
• Why the A and S keys - two of the most common characters in English text - are on my two weakest fingers, I’ll never know. The pinkies get more of a workout than usual because they’re handling all of the keys on the edge of the keyboard, too. Placing these two common letters on already weak and overutilized fingers is probably the biggest flaw in the Dvorak keyboard that I’ve found.
• Dvorak is lovely for English text. It’s just a great feeling to feel the letters whiz by with so little effort. However, a lot of my typing load is not English text. It’s Linux terminal navigation, C and Python source code, all of which intentionally discard vowels in exchange for brevity. This makes Dvorak’s plan to alternate hands work very poorly - you’re just not typing anything on the left hand.
• Some common Unix commands are absolutely worst-case scenarios on Dvorak. Take the command ‘ls -l’, which I would type dozens of times per day. On Dvorak, L, S and hyphen are all on the right pinkie. If you’re using a standard keyboard, so is the Enter key. It’s really, really unpleasant to type ‘correctly’.
• Naturally, position-based key bindings don’t work: Vim’s HJKL, WASD for games, Ctrl-Z/X/C/V for word processing.

I really tried to make Dvorak work. I gave it a month. At the end of the month, I switched back to QWERTY. It’s a fantastic layout for cranking out lots of English text, but that’s not my use case.

On the Kinesis keyboard

• The up and down arrows are backwards compared with the Vim convention. That’s relatively easy to remap on the keyboard.
• The macros are rather buggy. Actually, scratch that. The keyboard’s firmware is rubbish. I suspect that either the authors had never programmed on a microcontroller before or were not trained as programmers to begin with. I can crash the keyboard in two fairly common scenarios. That said, the crap firmware is not a reason to not buy the keyboard. It’s perfectly usable despite its flaws.
• I opened it up to look inside. It’s pretty typical of small-scale electronics manufacturing: no surface-mount parts, revisions hot-glued to the case, off-the-shelf components. It’s not badly made - it feels very solid for its weight - but it does have a few rough edges which might surprise you if you’re expecting a mass-produced product.
• It’s ripe for hacking. The main controller is an Atmel AT89S series microcontroller. The macro RAM is on standard serial EEPROMs. There’s even a socket for a second one (the Advantage Pro upgrade). Apparently the firmware can be changed over the PS2 or USB port, but Kinesis didn’t seem to willing to send it to me when I mentioned I wanted to modify it.

Where to from here?

I still want to improve the keyboard layout. The Kinesis makes typing less painful, but some of my pains appear to be linked to the QWERTY layout. And the feeling of effortlessly flying through English text with Dvorak was just amazing.

To come up with a better keyboard layout, I want to log my keystrokes for a month. Each keystroke will be tagged with the time and the active process. The process lets me figure out whether the keystroke intent was a letter or a position. I can also detect errors by tracking Backspace presses. With that information, I can determine exactly which keystrokes or combinations are the most common for me.

In addition, I want a ‘trainer’ - a program that will prompt me with an arbitrary series of keystrokes and time how long it takes me to hit them. This will give me information on how strong and fast my fingers are and if any of them are particularly error-prone. From that, I can generate a map of the keyboard, each key associated with a ‘performance’ score. Combining the two datasets, I can then come up with an ideal keymap for me, given my typical usage patterns and my own brain-keyboard performance data.

I’d also like to integrate information on common digraphs, but I’m not sure how best to use them. I’m not sure that Dvorak’s assertion that alternating hands is the best thing to do. A common case for me is the ‘chording’ I mentioned previously, where a single hand can hit a sequence of keys very rapidly. The timing is simpler - I arrange my hand correctly on the keys, then use the individual fingers to press them in sequence. Of course, this sounds like the sort of thing that might cause tendon damage. But it’s fast.

There’s more discussion on chording and performance here.

Development progress

I’d say I’m about 30% of the way to a SyncDroid release. It’s frighteningly small, but I know that boring stuff like testing, installers, licensing, graphics and documentation is going to eat up a lot of time.

I keep detailed lists for this stuff. Some people say it’s obsessive-compulsive; I say it’s damned well essential because there is just so much to do. If I don’t write it down, I’ll keep worrying about it until it’s done, and that’s a huge energy sink.

I’m at a point in development where I can finally say I understand how the whole application is going to fit together. It’s very reassuring - there was always this uncertainty as to how the different modules were going to connect (in particular, how asynchronous events were going to move around.) Now that I have a model I’m happy with, it’s just a matter of ‘colouring in’ what’s left.

Typing

I’ve had some level of hand pain for probably a decade. It’s mostly been quite ignorable and always went away after a few days. I blame it on the 1996-1999 period where I trained on Quake variants for anything up to 8 hours per day instead of studying.

Lately, though, it’s been getting worse. Much of this is that I’m doing a lot more typing now that I’m not rotting away in meetings. Short-term, I did an eleven-hour-long adrenaline-fueled hack-fest (on client work, no less). I didn’t type much for the following two days due to the pain. After some encouragement from my officemates, I bought a Kinesis keyboard.

I’ve seen them before, but ignored them due to the high price. Now, with the low US dollar, my improved financial situation, increased frequency of pain and likely expense of surgery, it seems like a bargain. I’m learning Dvorak on it too. It’s intensely frustrating to not be able to bang out 100+WPM without thinking about it, but so far I’m seeing major reductions in pain. I see daily improvements in speed and accuracy, so I’ll almost certainly stick it out.

Logos and icons

I can imagine what I want in graphics, but can’t express it (in words or images). I need some icons for SyncDroid and some logos for the website(s), so I went to a <$40 logo design place.

Lesson 1. If you say you want the logo to look like something, they’ll give you exactly that, copyright law be damned.

Lesson 2. If you let them pick a colour scheme, they’ll pick something ugly.

Lesson 3. If you specify colours, they’ll use them in the most ugly way possible.

Lesson 4. If you specify “simple, no detail”, they’ll add superfluous detail that would suck even if you had the resolution to display it correctly.

I’ve got ‘unlimited revisions’, but that doesn’t mean anything if they can’t design anything that looks good. So I walked a few doors down to an art supplies store, had a nice chat with the hottie at the counter, and bought a nice soft clutch pencil. I’ve sketched out some designs that don’t completely suck. I did have a quick go at implementing them in Inkscape but lost patience quickly.

I’ve looked at a lot of icon sets on the web and can see a pretty strong correlation between price and quality. Damnit.

Tuba

There’s a guy outside playing Mario Brothers and Star Wars tunes on the street corner. Newtown is by far the most fun location that I’ve ever worked in.

I’m here! I’m alive!

I’ve been doing a lot of client work lately, more because people want work done rather than because I want money.

Since the last update, I’ve made great progress on SyncDroid. After thinking about this thread on the Business of Software forum, I’ve decided to split SyncDroid into several different, smaller products - one for USB drives, one for network sync, one for SaaS Internet backup, and so on. It also helps me to focus - I’m working on a smaller product with a more clearly defined feature set rather than this great big ‘do everything’ product. And it helps me to focus the marketing message and differentiate (people like specific things that fit them well rather than generic things that happen to fit them). So I’m very happy with my decision to split into a whole product line.

UI work continues. I’ve got the scope of the project worked out and can finally say that I understand the project rather that just focusing on little bits of it. I can also confidently say this is where my product fits in the world and this is why you should pay money for SyncDroid instead of Unison. It’s a nice feeling. Very confidence-inspiring.
I’m amazed at how far off my original effort estimation was. I wrote most of the sync engine and figured I was about halfway. Now I can say that the sync engine was maybe 2% of the total effort required. UI’s are hard. Marketing is time-consuming. I haven’t started doing any serious cross-platform work yet, either - I’m doing a Linux release first.

I’ve been thinking about the different platforms a bit, too. My feelings are that:

• Linux is easy and fun to develop for and you’ll get good feedback, but nobody will actually buy your stuff
• Windows has a huge user base, but they’re relatively clueless
• OSX has a fairly intelligent user base and they’re trained to expect to pay for good software, but there’s just not that many of them

OSX has the additional problem that in order to develop for it, you have to buy a Mac. This becomes a simple economic problem: if I spend AUD$850 on a Mac Mini and $2000 worth of labour on an OSX version, will I sell enough Mac licenses to make the money back?

After spending years looking at everything through purely economic terms, I now believe that a decision based on such terms is an oversimplification. If I was going for (legal) profit maximization, my strategy would be:

• Go to my contract clients, start billing a lot of hours, and use my position as sole developer/provider to leverage a large $/hour rate increase
• Move back into my parent’s place
• Stop spending money on things like petrol and coffee

The economic decision-making model ignores the fact that my life would suck if I followed it through to its logical conclusion. I would be miserable. Starting this MicroISV has been one of my best decisions in recent memory just because of the quality-of-life improvements that it’s made. I’m not making anywhere near as much money, but I’m a lot happier. And so, I can confidently say things like I will keep doing this even if I’m just breaking even. I can happily forfeit a large income and go for a riskier, cheaper existence, because my life is that much more enjoyable.

And so, Buying A Mac. I’m going to wait. I’m going to get a Linux and a Windows release out there, gather some feedback, measure the size of the response. If nobody is interested in SyncDroid, I’m better off cutting my losses immediately.

I was going to end the post there, but it’s such a down note. Running a MicroISV rocks. Hard. Down with The Man!