When I worked on my high school paper, I spent many hours working on the newsroom Macs, but I had a Windows machine at home, and I was pretty comfortable going back and forth. I noticed a few people who had more trouble: Windows users who would swear a Mac would crash as soon as they touched it, and symmetrically, Mac users who if using a Windows machine would crash it right away.

This was in the bad old days of Windows 3.1 and System 7, before MMUs and memory protection, and crashing neither OS was particularly hard or uncommon. Still, there seemed to be something to this cross-platform-and-crash phenomenon. A friend and I came up with a half-serious theory that these crashes had to do with natural rhythms of human-user activity — mouse and keyboard input would arrive and trigger interrupts and Mac OS and Windows might have different preferences for the intervals between these interrupts where certain rhythms would trigger race conditions just so, and kaboom, while other rhythms would dance around the same bugs without triggering them. And, the theory went, people would subconsciously learn what sequences of inputs would crash their computer, learn to avoid those patterns, and compute along happily on whichever platform they’d learned to get along with: then upon trying the other platform, kaboom again.

It was with this in mind that I read John Moltz’s impressions of Microsoft Surface. He’s generally pretty fair about giving it a chance, but then he dings it for screen-rotation flakiness:

Once when I set it down the screen orientation was upside down and it didn’t seem to realize it. I had to pick it up and turn it over and then back again to get it the right way.

I’m sure that happened, but hey, every Apple device with automatic rotation (iPhone, iPod Touch, iPad) I’ve ever owned occasionally does that too. Come to think of it, so do the various digital cameras I’ve owned (from Canon, Nikon and Panasonic). I can’t think of anything that has a similar feature that I haven’t seen flake out that way.

But maybe if I could just learn to hold it right…

Vanessa and I went on a weekend trip to Orange County with Dominic and wanted to relax by watching a movie when he was napping, and the easiest vehicle for this seemed to be our iPad. What could go wrong? Lots, it turns out.

I cruised over to the increasingly inaccurately named iTunes store and found Safety Not Guaranteed which looked interesting so clicked Buy, waited for it to start downloading, watched the progress bar long enough to verify it looked like it would take about 40 minutes to download, and took Dominic out for a walk so he’d fall asleep, with the idea that in 40 minutes we’d have a sleeping baby and a watchable movie.

Dominic behaved as I’d hoped; the iPad didn’t; after about an hour I grabbed the iPad to find absolutely no sign of the movie. The Videos app said “no videos, but you can cruise over to the store to buy one”. The iTunes store app didn’t have any downloads in progress, and going back to the Purchases tab, it had the cloud icon next to our movie. Upon clicking that, it told me it can’t download it because there’s not enough storage available.

So apparently when I first purchased the movie, there was enough storage to download it, but then something went wrong with the download, it silently canceled the download without showing me any error message whatsoever, and now there’s not enough space for it? That’s hard to explain, but whatever, there’s nothing to do but try to clear space and download it again. I go to Settings: General: Usage and look at the breakdown of what’s using how much space, and the individual app usage doesn’t add up to anywhere close to what it says is used, but I don’t have many options here. So I deleted Infinity Blade to free up a gigabyte, invoked the download, and waited another 40 minutes while it downloaded the movie again. At the end of the download, the iPad spent another 5 minutes saying “processing”, then gave me a generic error message, and kicked me back to the state where the movie is not available for playback but is available for download. And again trying to download it says there’s no space available, and going back to Settings: General: Usage, there’s 2.7GB (the size of the movie) less than there had been when I started the download. Something very weird is going on. I can’t fix this problem on the iPad itself (so much for the post-PC era: you need a PC to configure or troubleshoot your post-PC device!) and now it’s been 2 hours and we don’t have time to watch a movie anyway.

When I got back home from the trip, I plugged the iPad into my Mac and launched iTunes and iTunes told me that there’s 10GB of “other” content. Given that this is the 16GB model (whose actual capacity is 13.8GB; go figure), that doesn’t leave room for very many 2.7GB movie downloads. That also explains why the total usage was so much higher than the sum of the space used by each app. But it doesn’t at all help explain what “other” actually means; the Internet isn’t much help here either. Trying to get this to go away or at least explain itself, I tried syncing the iPad with the Mac a couple times. The first thing I got for my trouble was that iTunes decided I had 7.1GB of video on the iPad, and now I’m over capacity by 6GB (this without me having added any additional content to the iPad, actually). The second thing I got for my trouble was that iTunes decided to sync that video content back to the Mac, which took all night, and resulted in 70GB of corrupt data: it created 3 .m4v files which were actually directories containing what looks like the entire iPad filesystem, 2 of which were named after the movie that caused/revealed all this trouble (Safety Not Guaranteed) and the 3rd of which was an episode of Breaking Bad which I didn’t realize was on the iPad in the first place. I’ll repeat: what got synced back to my computer in each case was not a valid .m4v file, but an entire directory structure whose top level ended in “.m4v”; one of these was 10GB, one was 20GB, and one was 40GB. Clearly, these weren’t actually on the iPad, and no, my Mac didn’t really like the resulting 70GB of garbage data either. (Something created what should have been .m4v files on the iPad as symbolic links to /, maybe? The entire directory listing is here if anyone finds this edifying.)

Back to reality. Apparently the fix for this runaway “Other” usage is to back up the iPad, then restore it from backup. So I did this, and after half an hour, I had a working iPad with 2.2GB of apps, 3.3GB of “other”, and 7.3GB free. That’s still not very satisfying, though — what is this 3.3GB and how do I get it back?

So then I tried completely erasing and restoring the iPad (restoring the firmware, not just the user data, which is a similarly named but differently implemented operation in iTunes). This worked a little too well — the iPad came back with only 0.62GB of “other” and 13.2GB free, because it has no apps installed. Um, that’s partially my fault, because I’d turned off syncing of apps in iTunes (in my experience, leaving app syncing enabled makes syncing take much longer and sometimes forever, and it shouldn’t be necessary in the post-PC mode where you manage the device on the device itself); I don’t find it obvious, but apparently if you disable the syncing of apps in iTunes, that also means that local restores via iTunes don’t restore apps either.

But oh whatever, I have a recent iCloud backup which should have all the apps backed up. I do another complete erase of the iPad, and this time when it boots up, instead of driving the restore from iTunes on the Mac, I walk through the on-device setup and select my iCloud backup. It chews for a while, reboots, and then gets stuck forever at the boot screen (Apple logo with progress bar). Apparently I’m not the only one with this problem either. I tried again and got the same result.

So apparently the iCloud backup is corrupt and can’t be used; the local backup works but leaves me with extra work to do.

In the big picture, this is just another internet rant about a bug which happened to one person and probably isn’t generally applicable, and while I can describe it as 6 separate problems (iPad aborts my download without showing an error message, iPad downloads movie to bit bucket and won’t play it back, iPad runs out of space, iTunes can’t sanely describe space usage, iTunes syncs garbage from iPad back to my computer, iCloud backup is corrupt) it’s likely all one root cause, where some data structure got corrupted and caused the rest of the failures downstream from there. And computers are complicated, and not that expensive given the complexity, and hard and expensive to get right.

But what’s frustrating about this is that Apple’s moved the openness needle pretty far back toward walled garden, in the name of preventing problems. Problems like malware, yes, but also problems due to misconfigurations, and problems due to third party software that causes finger-pointing instead of solutions. This was all first-party software; Apple content in Apple software on an Apple device synced and controlled from another Apple device. No jailbreaking, and I didn’t even commit the minor sin of trying to run iTunes on Windows. This is the case that we should be getting right; that’s the benefit that we’re paying the walled-garden cost for.

Apple is often applauded for bringing us technology that “just works”, but I’ve been having more and more experiences lately that are pretty far from “just works”. The moral of the story is that Apple’s devices are computers just like other computers; they’re still a stack of fiendishly complex hardware and software components; also with iOS in version 6 and iTunes in version 10.7 there’s a lot of untestable legacy cruft. “Just works” sounds nice, but we’re not really there yet, or maybe we’ve been there but don’t know how to stay there for long.

I read two extensive reviews of Microsoft’s new Surface tablets today.

Peter Bright, writing for Ars Technica:

They’re covered in polyurethane that should prove hard-wearing and water resistant (I was told that they should withstand scouring pads, though I admit I’m too chicken to try—we had to promise to return our review units unharmed)

Mathew Honan, writing for Wired:

And it’s quite sturdy, to boot. In Microsoft’s lab, we saw Surfaces that had been opened and closed hundreds of thousands of times. They still snapped open and shut like they were new. We wanted to see how easy it was to break one. It’s very possible, but you have to really try. We did manage to break off the kickstand by gradually leaning onto it, but I had to put nearly my full weight onto the tablet before the kickstand snapped off.

and

Microsoft showed us a drop test by letting a Surface fall from a chest-high height. I duplicated this by literally throwing it at the floor, but to no ill effect.

All in the name of science, of course. I doubt Bright could have gotten in that much trouble for taking a Brillo pad to his Touch Cover. Oh well.

So, it’s come to this:

I waited a week after iOS 6 was released, asked my Facebook friends what they thought of it, and within hours got 6 responses that all said not to upgrade if I care about maps. Well, I do care about maps, but how bad can it be? I wanted to see for myself, but not shoot myself in the foot on the phone I depend on daily, so I upgraded my iPad instead.

Actually, the new 3D view is very cool where the data is good, and here in San Francisco the data is good. Zooming around my neighborhood is fun and looks great; that’s a real win. In fact the maps overall look great, and the zooming is much smoother than what we’re used to with the Google-powered maps in previous versions of iOS. How about the underlying data? Um, not so good. Roads are (in my experience) where they belong, but addresses and landmarks and businesses are often not. I don’t know why anyone is surprised by this; it’s a very hard problem, and Google has worked incredibly hard (and thrown a lot of money, human time, and computer time at acquiring and massaging the data) to get their maps to the point we all take for granted now. The world is very big.

That doesn’t mean it’s pleasant when I search for an address a mile from where I’m standing and get a search result for a similarly named street 1000 miles away in Texas. That’s what happened to me the first time I tried actually using Maps for real, i.e. to find something whose location I didn’t already know. I also tried plugging in a few locations I remember from our trip to Ireland, where iOS 5 maps (i.e. Google maps) got us everywhere we needed to be, and Apple maps didn’t know them; that trip would have been much harder.

But luckily, the built-in Maps app powered by Apple’s fledgling data set isn’t your only choice, and that’s what my screen shot above reflects. I tried the various alternatives shown here before upgrading, and came to roughly the same conclusions as this Ars Technica article: Waze is neat for navigation but not a replacement map; MapQuest is pretty unpleasant to use; the Google Maps website actually works just fine; upon confirming that last, I decided it was fine to upgrade.

More surprising to me is that it’s Nokia who has the second-best map experience — both data quality and interaction design — after Google. (Maybe I shouldn’t have been surprised. I remember reading about Nokia’s Mobile Millenium project and thinking that while their approach seemed more like academia than industry and Google is probably leaving them in the dust in terms of speed of deploying new features, their very thoughtful approach especially with respect to privacy is commendable. Clearly, they’ve been taking this seriously.) I tried, again, punching in some of the hotels we stayed in in Ireland, which Google could locate and Apple couldn’t, and Nokia found them just fine. That test involves just a few data points, but is in keeping with a general reputation that only Google and Nokia have decent place data outside the US.

So the obvious conclusion: Apple should just buy Nokia.

2006: The first RAID/storage appliance I bought was the original Infrant ReadyNAS NV. It was medium-expensive for its time, well-built, and at first I liked it a lot. However a few missteps by Infrant (one about spending years stonewalling user requests for ssh access while sitting on a MAC-address-based backdoor password, one about essentially a product recall of the fan which if you read between the lines just meant they’d installed it backwards and customers who want proper cooling should flip it around but it will get louder), one power supply failure, and some mediocre experiences with their generally well regarded tech support around buggy CIFS and AFP coexistence, then the fact that Infrant sold itself to Netgear (the conventional wisdom on which had it that there goes their tech support, although I didn’t notice any difference, and their jedi council is still at it today), combined with extremely middling performance, meant when I needed more space, I went elsewhere. (As you’ll see, in hindsight this feels misguided; 6 years later and the little ReadyNAS NV is still humming along fine, Infrant continues to support it with software updates, and it’s been far less frustrating and more useful than my next couple purchases.)

2008: Elsewhere meant Promise; I bought an NS4300N. It offered a fairly similar feature set to the ReadyNAS, but was half the price and slightly faster. It was also of notably lower build quality, a lot more rickety and a lot louder. Over time the firmware situation got sketchier; the product stopped evolving; I moved it to a friend’s house to use as offsite backup, then slowly stopped using it; now it’s gathering dust.

2010: Soured on the consumer-priced NAS market mostly due to poor performance (Infrant and Promise had both claimed speeds of around 30MB/second but that seems to involve way too many arcane dances and incantations involving jumbo frames and disabling journaling and other bad ideas, even for me; configured sanely, I see a reliable 10MB/second from the ReadyNAS, which is overkill for music storage and backup and frustrating for anything else, so guess what? I use it for music storage and backup), and still dreaming of fast centralized storage for photos and videos and home directories and data I care about, I tried a DAS solution next. The Promise DS4600, cabled via Firewire 800 to a Mac Mini which is on all the time anyways. (The decision calculus that lead to my original NAS purchase in 2006 involved servers that drew 100+ watts and marketing claims that the consumer NAS units performed fine and had reasonable AFP implementations; here in 2010 I was betting that I’d get better performance and correctness from the real Apple AFP server on a real (Intel Core 2 Duo) CPU and now that only draws 10 watts.) In retrospect, this was the worst of the bunch. On a good day it reads and writes at 50MB/sec, but on the much more common bad day it reads and writes at less than 1 MB/sec; as originally shipped it would somehow cause the host computer to slow to a crawl and need rebooting every few days, spending all its time in unbillable kernel space (how a Firewire device accomplishes this, I have no idea); a firmware update eventually fixed that, but to this day years later, performance is unpredictable and generally slow, especially for small random writes, and if I reboot the host computer, the DS4600 array won’t be visible until several minutes after boot. Also twice the array thought there was no volume configured (and it came back up reboot, but that’s still pretty scary, and meant I had to stop using this for anything important).

Oh, and the configuration app for Mac OS that ships with the DS4600 creates 1024 threads (for what unholy purpose, I have no idea) and causes 16 identical copies of the mouse cursor to appear onscreen, offset vertically by 64 pixels. Keep in mind that while I didn’t buy this from Apple, this is a product that Apple actually stocked and sold specifically for use with Macs, so I shouldn’t have been that far off the beaten path here. (This is tied with another Promise software disaster inflicted on me a few years ago for reasons not to keep buying their products. Their hardware seems to perform well for a low price, but riddled with software/firmware bugs as it is, no more for me. The other case I speak of: I bought a low-end 2-port SATA RAID controller for a Windows PC in 2006. All configuration for this PCI card was carried out by a Java app with an in-browser interface, meaning you need a Java VM and a web browser to twiddle a few hardware registers on a PCI card. Also, the installer failed to create any front door to invoke the web browser, so I had to run tcpview to figure out what port it was listening on to invoke it. All this for something I only had to use once, for one-shot configuration. The RAID controller worked fine for years, but what a horrible out-of-box experience. The resulting web UI for the PCI card and the standalone NS4300 RAID appliance was almost identical, I can almost cross my eyes far enough to envision a world where it made sense to do write-once-limp-anywhere for the configuration UI, but not quite, and in no case does that excuse making me run tcpview to figure out where to point the browser. Though to bring this full circle, this division of labor between the embedded software development and real-OS software development, or lack thereof, probably explains the 1024 threads the “SmartNAVI” app creates, somehow.)

2012: Thoroughly frustrated with the DS4600 and its foibles, lack of reliability, and unpredictable speed. And we’re shooting lots of video of Dominic, so I need somewhere big and fast to store video files. DAS no longer sounds like a good idea, largely because how am I going to connect it? My current home server, the Mac Mini, has only USB2 and Firewire 800 and those are no longer considered fast. Apple continues to willfully ignore eSATA; their anointed storage bus is Thunderbolt which is (still) a tiny expensive low-volume ecosystem which tends to point back to Promise, and which is not available on any of my current machines anyway; USB3 is starting to creep into their machines but as of October 2012 is still inexplicably only available on portables; there’s just no DAS array that I could plausibly connect to my current server and a future server and expect good performance. Now you’re probably wondering why I’m talking about Apple hardware anyway; clearly they have no desire to let us do external storage anywhere near the price- performance of the PC world; I should just build my own box and run Linux or Nexenta and soak in the ZFS goodness and… no thanks. I’m crazy enough to build my own networking hardware, but don’t want to deal with storage hardware.

So…. back to NAS? The last day the DS4600 thoroughly irked me, a few hours later Ars Technica ran this gushing review of the Synology DS-412+. Sounds good to me. I ordered one (actually the 5-bay DS-1512+, which cost hardly any more) and am setting it up now. Wish me luck. In first impressions, the hardware is well built, it runs quieter than either the ReadyNAS NV or the DS4600 that it’s meant to replace, and I like that they gave some thought to fan replacement and memory upgrades. The browser UI is a little too flashy for its own good, but is impressive in its own way. I haven’t done any real performance testing yet. Wish me luck.

Yeah I know, this is already well known and I’m just piling on and complaining. Still, it’s annoying and worth noting. Airlines keep adding arbitrary fees and reducing service quality. Boo. This is worse for us because now we’re flying with Dominic as a lap infant and the airlines are especially poorly optimized for this, at really every level. You can’t buy tickets the easy way, you can’t check in the easy way, and everything gets harder.

A few recent experiences that sucked:

Southwest: they don’t charge for a lap infant and are generally friendly, which is why we keep flying them. But if you have a lap infant, there’s no way to add him to the reservation, but you do need to get a boarding pass, which can only be done at the full-service checkin line, which is the longest and slowest one. They’ve increasingly incented people to check in online and either skip the front-of-airport counter if not checking luggage, or use the faster “bag drop only” counter if checking luggage, and only people who never travel and don’t realize any of this, or people with unticketed infants, end up in the longest slowest understaffed full-service checkin line that the rest of their policies are actively trying to drive us away from. Then, we need to gate check Dominic’s car seat, which can only be done at the gate counter. So every flight involves visiting both the front checkin counter and the gate checkin counter. This doesn’t waste just our time, but also that of the Southwest employees and everyone behind us in line, and really nobody benefits.

Aer Lingus: on our trip to Ireland, we booked our international travel from the US into Dublin and back from London, and left undecided how and when we’d get from Ireland back to London. We eventually booked an Aer Lingus flight from Waterford to London, which was easy enough (and even let us book a ticket for Dominic with no hassle which was a pleasant surprise), but booking this ticket on aerlingus.com also had an unpleasant surprise: they quoted the price in euros right up till the final “buy” button, when they converted it to dollars, at a rate about 8% worse than the going exchange rate. No way to opt out of this. I’d have been happy if they charged me in dollars at the current exchange rate, and I would have been fine if they’d charged me in euros and I paid the credit card company their foreign transaction fee (around 3%), and I would have been fine if they’d charged me in dollars at the current exchange rate and added a 3% currency conversion surcharge. But silently pretending the exchange rate was significantly worse than it is: that’s just shady.

United: trying to get to Mexico in December for a friend’s wedding, Kayak quoted us $830 roundtrip for 2 adult tickets on United, and there’s no way to add a lap infant through Kayak. I tried searching for the same flights on United’s website, where we could add a lap infant, and it came to $915, the extra $85 breaking down as $30 for the infant seat (fine) and another $55 in US taxes (not fine, because they’re repeats of taxes already built into our fare, and basically charging the same amount of tax for Dominic as for me and Vanessa put together). I don’t know where the tax money is going or whose fault this is, but it sure doesn’t seem right. So I called United to ask whether I should add Dominic to the reservation up front and pay those taxes or book without him and add him to the reservation after the fact, and they couldn’t answer the tax question but said the website won’t let me book a lap infant and I should definitely book without him then call back to add him. Even though that didn’t seem true, I was hoping the taxes would resolve themselves more sanely that way, so that’s what I did. Imagine my annoyance when I call back to add Dominic to our reservation and am told they will charge me $30 for his ticket, $55 in taxes which they can’t explain but they are what they are, and $25 in reservation fees for not booking online, which they now charge even for transactions which the website can’t do. (And while the website seemed willing to book the lap infant as part of the reservation, both customer service agents I talked to told me they don’t think it works, and that even if it doesn’t work, that isn’t cause to waive the $25 fee.) Fail.

On a less whiny and more humorous note, I’ll leave you with this. C’mon, airlines, we just want to pay a fair price for a fair service, no tricks or traps.

MG Siegler, Apple’s Magic Is In The Turn, Not The Prestige:

So while some may find Apple’s trick old hat now, no one else has figured out how to pull it off — except for the company doing a mediocre copy of the trick. I’d argue it’s because everyone is focusing on The Pledge and The Prestige, but Apple is the only one focusing on The Turn.

Also, I’ve been meaning to see that movie.

Apple has been dabbling with resolution independence in Mac OS since 2005; in 2011 they shipped double-resolution “Retina” displays in iPhone and iPad, and in that year’s Mac OS release (10.7 “Lion”), renamed the tentative resolution-independence support “HiDPI” and capped it to a 2x scaling factor, hinting strongly at the future release of Retina Macs. In 2012, they released the Retina MacBook Pro and soon thereafter, Mac OS 10.8 “Mountain Lion”, bringing the HiDPI modes out from behind the developer tools and debug flags.

My question was whether Apple would ever make Mac OS usable for media-watching purposes, connected to an HDTV, aka an HTPC. Connected to our living room 1080P LCD TV, we have an Apple TV, an Xbox 360, and a Mac Mini; the Mac is there as an HTPC basically to handle whatever the dedicated media devices can’t; the list of things the ATV and Xbox can’t handle has been shrinking over time, but still. The pain point in this arrangement has been that if I set the Mac Mini to full 1080P output, text onscreen is too small to read. If I set it to 720P, text is borderline readable but still on the small side, and images are scaled by 1.5X in the TV, so there’s some loss of picture quality once we get into a movie.

This scenario seems a pretty good match for the HiDPI modes — specifically 960x540 HiDPI, meaning text and UI elements are rendered for a virtual 960x540 resolution but the display itself is running at its native 1080P (1920x1080, exactly double the virtual resolution in each direction) and images get to address every pixel. (It seems faintly ridiculous to me that Mac OS had to wait until 2012 to offer any kind of control over the size of UI elements, and that after waiting until 2012 it arrives in such limited form; Microsoft Windows has offered control over UI element text size since forever; in practice apps with custom skins would often misrender some controls if you used custom text sizes, and I suppose Apple was trying to limit exposure to that problem by exposing such a little and late form of resolution independence. Did it work? Read on….)

Note that by Apple’s definition, a 32” 1080P display from a distance of 8 feet, across our living room, very much qualifies as a “Retina” display. You can’t see individual pixels, and both text and images ought to look pretty darn great.

In my experience with the “960x640 (HiDPI)” mode in both Mac OS 10.7 and 10.8, by forcing the 2X scaling factor and automatically pixel-doubling unaware apps, Apple has successfully avoided the problem with both Windows’ themed UI font sizes and Mac OS 10.4-10.6 scaling factors where some UI elements would be clipped to the wrong size. On the other hand, by giving me 960x540 as the only resolution that both addresses every physical pixel and has readably larger text, they’ve introduced a new problem: nobody cares about making apps actually fit on screen sizes this small. (Typically, the archaic and once- grand XGA, 1024x768, resolution is taken as the smallest possible, and indeed Mac OS warns you if you select a resolution smaller than that in either direction that apps may not fit.)

The problem starts with Apple’s own apps; here’s Software Update (from 10.7) on the 960x540 screen. (A couple notes here. First, the sad thing is this app is not design-intensive and does not require any special UI elements; the window itself and the individual panes are resizable; there’s no reason it couldn’t run perfectly well at 960x540. Second, in 10.8 this app has been replaced with the Mac App Store app, which does a better, but still not good, job of fitting at 960x540.)

Note that the initial window size (above) doesn’t even try to use all the available screen height; note also that enlarging it (below) helps not at all with the real problem.

Hoping that 10.8 (the first OS X release conceived and shipped for Retina Macs) might fix this oversight, I upgraded the OS on this machine. The initial results are not pretty. The first thing you see when launching the Mountain Lion installer:

Such crisply rendered text! (Click through for a high-res version.) Nice large cougar image! Such beautiful use of whitespace! But… where’s that promised Continue button?

(I cheated, changed display resolution to reveal the Continue button, pressed it, then changed resolution back to 960x540 for the next screenshot.) Again, the same problem.

OK, you say, but those images are still from OS X 10.7. It’s an app designed for and as part of 10.8, to be sure, but maybe 10.8 itself will be better? Reboot into the new OS install, and:

Rats. Foiled again. At least the clickable buttons are onscreen, but the screen was obviously cropped, and not intentionally.

No, Apple, thank you. (And a special extra thanks for not starting the screen sharing daemon until after I click these two right arrows from the local display, preventing people from upgrading their Mac OS installs remotely. You can do the rest over screen sharing / remote desktop, but you need a local display and mouse to get past these two screens.)

The real problem here isn’t that TV-as-Retina-display-with-HiDPI couldn’t work, it’s that apparently nobody cares. Nobody cares about screens smaller than 1024x768 because nobody’s shipping those any more; nobody cares about making 960x540 work as a special case for TV use, because you’re just supposed to buy an Apple TV instead of hooking a Mac up to your TV, I guess.

So close, but yet so far.

Since Vanessa and I have fairly similar cars, one gasoline and one diesel, I found it instructive to compare their observed fuel efficiency now that we’ve put a few thousand miles on each.

One is a 2012 Audi A4 wagon (4-cylinder 2.0 liter turbocharged gasoline engine). The other is a 2011 VW Golf TDI (4-cylinder 2.0 liter turbocharged diesel engine).

For both of these, I’ll list real-world numbers for the speeds I actually drive, averaged across a bunch of real driving. (I wish it were easy to get numbers like these before buying; the EPA test circuit isn’t all that reflective of our real world driving). I’ll list the fuel mileage for 3 cases which I find instructive:

• best-case I’ve seen (long-distance highway driving)
• a typical commute from San Francisco to Palo Alto (37 miles, mostly highway)
• the average over the entire time we’ve had the car (a mix of city driving, medium-distance commuting, and longer road trips).

One thing I already knew: the diesel engine is more efficient than the gasoline engine. Another thing I already knew: highway mileage is better than city mileage (excluding hybrids with regenerative braking, which these cars are not). And one thing I didn’t know, but learned from reading the Audi owner’s manual and confirmed by real experience: short trips get worse mileage than long trips, because mileage is worse before the catalytic converter warms up.

On to the numbers (all of which come from the cars’ own trip computers).

Audi A4 (gasoline): average over our entire mix of driving is 25 mpg. Typical mileage on the commute is 29 mpg. Best mileage I’ve seen for an extended period of highway driving is 33 mpg.

VW Golf TDI (diesel): average over our entire mix of driving is 37 mpg. Typical mileage on the commute is 44 mpg. Best mileage I’ve seen for an extended period of highway driving is 51 mpg.

Conclusions and observations:

• The Golf TDI definitely gets better mileage than the gasoline A4. About 50% better across the board, actually. (Though there are many variables other than the fuel type; the A4 is bigger and heavier, and has all wheel drive and an automatic transmission. On the flip side, bigger isn’t necessarily worse; it’s likely more aerodynamic. And Audi claims their 8-speed automatic transmission delivers equal or better fuel economy to the 6-speed stick.)
• The Golf TDI is more sensitive to speed than the A4; there’s more upside to driving at or below the speed limit.
• There’s an interesting comparison to be made, for both cars, between the “typical commute” and “best I’ve ever seen” numbers. That typical commute is 37 miles starting and ending with a few non-highway miles; the average mpg starts low (non-highway), climbs as I get on the highway, then climbs further as the catalytic converter warms up. By the time I arrive, the running-total average has pretty much stabilized, but only barely. And the going rate on Highway 280 is often around 75 mph, well above the optimum speed for maximum efficiency (closer to 60 mph for both these cars, as far as I can tell).
• The “best I’ve ever seen” results derived from getting on the highway and driving just over 60 mph for a longer distance — so the steady state average mileage is higher (because of lower speed) and contributes more to the final average (since most of the time was spent in the steady state after the catalytic converter warms up).
• Our observed around-town mileage for both cars is significantly worse than the EPA “city” estimate, because these trips tend to be very short, full of stop-and-go, and at suboptimal speeds. I think the EPA cycle tries to account for the latter 2 factors but not the first one. This is true for the diesel car as for the gasoline one, and is a further argument for electric vehicles for in-city driving.

Apple pretty much owns the tabletmarket today with the 10” iPad, but the Nexus 7 is here and good (in my opinion, the first viable answer to “why buy this instead of an iPad”), and meanwhile, rumors of a 7-ish-inch “iPad Mini” or “iPad Air” are heating up.

My family has both a Nexus 7 and an iPad, and I find myself using the Nexus 7 more than the iPad, for four reasons. This makes me think that the iPad Mini rumors are likely true. In a nutshell, while a 7” tablet can’t do everything a 10” tablet can do, the 7” tablet can do the things where I’d choose a tablet over a laptop or desktop, and it can often do those things better than the 10” tablet.

First, the reasons that limit what I do on a tablet vs a laptop or desktop:

1) the iPad (like all iOS and Android devices) doesn’t have user accounts, yet my wife and I share the iPad, and I mostly leave it lying around the house. So I haven’t given it credentials for all the individual apps and services I use — in some cases because Vanessa and I would both want to use them; in some cases because I’m not comfortable leaving that credential cached on a device I don’t keep close tabs on. So some of the apps I use the most on computers and my phone, and would potentially use on a tablet, I don’t use on the iPad: email, Facebook, Evernote. The ability to have user accounts or profiles (and have some apps available without unlocking an identity, and other apps behind an identity) would make a big difference here; I have to assume this will happen at some point.

2) typing on any of these onscreen keyboards is frustrating, so I tend to just wait till I get to a computer for serious input. This hit me when I was watching my son Dominic learning to crawl: at first he was perfectly happy to drag himself around on his belly with his arms, his legs trailing behind; as he got closer to real crawling, he started getting frustrated and impatient with the previously good-enough belly slide, because he had the yearning for real speed. So I feel about typing on a tablet: I can do it, and if I were trapped on another planet with no other means of self-expression I would surely use it, but if I have thoughts spanning more than a sentence or two, I’d rather just wait till I get back to a real keyboard to pour them out effortlessly. (And I realize I can use a separate keyboard with an iPad: it basically negates this disadvantage, but also the portability advantage vs a laptop.)

Then, the reasons the Nexus 7 succeeds handsomely at the modest task that is its charge:

3) the form factor. It’s significantly more portable, and significantly easier to hold in one hand; notably this makes it much better for reading on trains and reading in bed. Given the apps I’ve already stopped using (due to #1 and #2), these are the primary ways I want to use a tablet. The iPad’s bigger screen is an advantage, but not enough to make up for the reduced portability/holdability.

4) while Android as a whole doesn’t feel as fluid or natural as iOS, on both platforms the OS is designed to pretty much get out of the way once you’re in an app; that’s to say your real impression of the platform is driven by the apps. The main apps I use on the Nexus 7 are surprisingly good. The first- party Google apps (gmail, Google Reader, Google Maps) are better than on iOS; the builtin Chrome browser is very good; the third party apps I use the most (Kindle, Instapaper, Evernote, Facebook) are all available and good. So the overall impression the Nexus 7 leaves is very good.

So, iPad Mini…?

There are warts to Android — even with Jellybean, scrolling and touch interaction don’t feel as smooth as iOS always has. The back button is a disaster; you never know what it’s going to do. The app selection isn’t as wide as iOS, and even where equivalents are available, they’re often rougher. Plus, I’ve already paid for a fleet of iOS apps. While it’s a tradeoff with #4 above, I have a feeling an Apple 7” tablet could be even better than the Nexus 7.

What all this means is that given the way I use a tablet, I find the Nexus 7 does everything I want it to, far cheaper and more portably than the 10” iPad. I look forward to seeing what Apple does with the rumored iPad Mini (or iPad Air).