Think about this - when Android goes to tablets with Honeycomb, the same app binary will work with tablets and phones while automagically adjusting the UI to suit the available screen real estate. Now think how Apple did this with iPad.
Think about inter-app communication. Intents are another one of the great platform facilities. Your app can get things done at runtime without worrying about if there is an app installed to handle that intent - if there isn't one it will launch the Market app and suggest you to download the necessary app to handle that activity. The cloud Backup APIs and C2DM stuff is way ahead than what Apple has to offer today.
Then there are obviously the notifications. Ever heard any one complain about Apple's horrid notification system? Nope - they are all busy looking at UI animations :)
Not having to deal with memory management is added bonus.
There are a lot of things Android gets right. If they don't have as many rich animations as iOS it doesn't make a whole lot of difference to end users - having crappy notifications on other hand does affect end users negatively.
So all put together I think there are lot many pleasant things to say about Android as a development platform. Purely from an "emotional" or "UI" angle it may not be terrific but from an getting work done fast standpoint it surely is a great platform that also happens to look decent enough not to make you puke ;)
Obviously, the operating system is not going to be able to produce the same design decisions that would be made in light of actual experimentation with real users - it will be based on heuristics that will produce sub-optimal results for just about any combination of screen size and resolution other than the original. So I'll assume that you merely meant to assert that the scaling will produce results that are "good enough" in that normal users won't notice that the UI doesn't fit their tablet.
Even the claim that it will be "good enough" seems outlandish. Consider the kinds of scalable GUIs that exist on desktops: the important elements of the GUI (like toolbars, menus, and standard buttons) don't get re-sized. At best, they get wrapped or scrolled, either of which is instantly recognizable as a bad design. Most of the sizing flexibility on desktop apps is in the areas that show relatively static content that the user only interacts with passively, like in a web browser or video player, or in the case of many types of games, areas that you only interact with in small bits at a time. (Word processors alternate between the presentation of static content and the interaction (ie. editing) of one spot at a time.)
On a multitouch display with direct manipulation of the GUI, you will typically have far fewer opportunities for the easy scalability that web browsers and word processors enjoy. You have to know which UI elements need to maintain the same physical size, which ones need to maintain the same pixel dimensions, and what the limits are on the scalability of the other elements.
Take as an example the iPad email app. In landscape mode, you have a list of messages on the left (showing subjects, the first line of the body, and a few other pieces of info), and the body of the currently selected message shown on the right. The right pane takes up about 2/3 of the screen width.
If you were to scale this app from the 4:3 iPad display to a 16:9 tablet, how would the left/right split ratio have to change? It's complicated: the entries in the left-side list will have to maintain roughly the same physical height, because they are essentially buttons. You probably want the left pane to remain at least as many characters wide as in the original, so the physical width won't change by much, but you have some flexibility in font sizes. That means that the right pane showing the body of the email will probably be slightly wider and shorter on a 10" 16:9 tablet, but on a significantly smaller tablet, the left/right split might be 50/50.
All that is merely a first approximation of how that one app would have to scale. Could a computer derive that? How will the operating system know what the driving constraints are? The OS certainly can't know whether it is better on average to use the extra width of the 16:9 tablet to show more of the subject lines on the left or more of the body on the right.
It seems that the only way to make a scalable touch UI work is for the designer to bake in the constraint information, which will require testing at different sizes. This obviously doesn't save any work if you are only targeting two sizes. How many different sizes do you need to target before the rule-based scaling has a significant advantage over manually tuning the different sizes? Whatever that number, Android will exceed it and Apple probably won't. However, this completely ignores the need for more drastic design changes when the difference in sizes is very big, such as between the iPhone and the iPad.
And such discontinuities in the scaling hardly require that big a disparity: according to AnandTech's review of the Samsung Galaxy Tab (http://www.anandtech.com/show/4062/samsung-galaxy-tab-the-an...) which is a 7" 16:9 tablet, the best way to type on it is in portrait orientation using your thumbs, whereas on the iPad (9.7" 4:3) you are better off using both hands in landscape. On a 10" 16:9 tablet, you have even more width in landscape mode to accommodate typing with both hands, but how is the OS supposed to know that in portrait mode it is still too wide for two-thumb typing?
The only really good scalable UI engine I know of is LaTeX, which outputs completely static content, but can handle a wide range of paper sizes without the author needing to worry about margins, and all the font sizes can be scaled with a single change to the header of the source file. Even so, it doesn't know when the paper size is large enough that you should switch to a two-column format.
"[Fragments] allow apps to split their functionality on varying devices. In other words, the Gmail app on an Android phone will load your inbox on one page, then slide to a new screen to display an email. On an Android tablet, that very same app will display completely differently, loading the inbox in a left-hand pane and using most of the screen real estate for emails. Rubin stressed that Android engineers have thought hard about making sure apps are backwards compatible with phones but still run well on tablets. Rubin started to talk about how fragments will work on Google TV before switching topics--we’re guessing he was simply going to specify that some apps will fragment to take advantage of the TV platform as well."
Actually, that makes much, much more of a difference to end users than notifications systems or "intents" (both of which, I freely admit, are very awesome). Try a demo for five minutes with an Android phone and an iPhone. Even if the Android phone was more productive in the long run -- and I'm not sure it will be, but let's say that it is -- the pretty UI will win every time.
> the same app binary will work with tablets and phones while automagically adjusting the UI to suit the available screen real estate
Is that a good thing? I'm inclined to say that it's not. With a click of a button in the IDE, you can convert your iPhone app to an iPad app, having the OS scale up the UI for you -- it's just a horrible experience. It's not a tablet application. It's an application designed for phone usage patterns, with text columns too wide to read, and reasonably unusable.
The iPad Human Interface Guidelines say that "The differences between Mail on iPhone and Mail on iPad reflect the different user experiences of each device." You can use one binary for both interface idioms, but trying to use the same interface, or a slightly changed one, just ends up in a horrible app. I don't want to use a tablet with auto-scaled phone applications, and if that's what Honeycomb makes developer think is somehow a good thing, then shame on it for ruining the Android tablet experience.
> Ever heard any one complain about Apple's horrid notification system?
I hear complaints all the time. It is awful beyond belief. There is no way to deny that. So was the lack of multitasking about this time last year. That was fixed.
> Not having to deal with memory management is added bonus.
I develop for the iPhone, and I don't "deal" with memory management. I follow two rules: release what you own, don't release what you don't. (I actually like dislike garbage construction because it causes issues with destructors and non-memory resource management, but that's a different point.)
> So all put together I think there are lot many pleasant things to say about Android as a development platform.
I agree. There are. But I don't think it's inherently better than iOS, although the Java roots might make it much more familiar to a large section of the programming population (that, however, might also use some experience in a new type of dynamic language).
Also GC mostly is huge benefit when it is available - like everything it has its downside as you rightly pointed out, but I prefer that over having to remember to release memory in different code paths. Gives me more time to think about program logic. Plus you can always write native code in NDK when you want more control over object deallocation for performance critical code.
There's this magic thing called "autorelease".
Call it on an object, and it'll release itself (and its memory) sometime in the future, after the current call stack (probably sometime a few runloops ahead). Most things come pre-autoreleased for you, the only time you ever have to manage memory yourself is if you are storing it inside the instance variables of an object, at which point you just have to remember to add a line to the destructor to free it.
That may be wonderful for developers, but that will probably make for a horrible user experience.
There are no two apps in case of Android, only one App with no conditional coding required to support a tablet and a phone. The underlying framework will do the right thing automatically based on where it is running.
And it seems like no one on the Android team thinks that GPU acceleration of the UI is a key problem. http://code.google.com/p/android/issues/detail?id=6914
GC is far more likely cause for the jerkiness. The 2.2 and below GC implementation is not very latency friendly. They fixed that in 2.3 with concurrent GC with a target GC pause of 3ms. Looking at 2.3 on Nexus S - there are literally no user visible pauses / jerkiness - I just can't bog it down.
GPU acceleration may be useful in some specific cases and I am sure they will get there - but having to support bazillion devices that have different GPUs - that's going to take little longer.
How much market share does Android 2.3 have?
As a developer, I like Android. I really do. I'm pulling for it. But as someone who has helped develop a first class Android application, the hardware fragmentation is the largest headache I've had to deal with. Even pulling out all performance enhancing techniques, it is hard to deliver an experience on a G1 running Android 1.6.
I want Google to take a firmer stance around the hardware that powers their devices. Call it brand protection, a licensing requirement or just plain "we're better at tech than you." Please, just make the phones work.
Oddly enough, it's Microsoft doing that, not Google.
G1 is a bad example - it is too weak spec wise. Aging hardware problem is not something anyone can do anything to resolve. With Google the pace of obsolescence is faster than Apple's but that's a fair price to pay and not many common apps are demanding enough to care about hardware specs. There is always going to be faster hardware and specialized software that exploits it - it's not a problem, just evolution that you got to manage.
Also now a days most Android phones do have decent enough hardware to not cause you the G1 class of headaches. So there isn't much to worry if you target 2.1+ phones which gives you 85% coverage.
The only solution for the developers of that uncommonly demanding app is to have minimum specs requirement for the app. And I have heard that Honeycomb will come with base minimum hardware requirements - but that may be a moot point as already most Android devices have decent hardware now a days.
The whole point of the article is that the subjective matters -- literally to the tune of Billions.
So it is possible today for people to buy Android devices and not notice any out of the ordinary jerkiness.
Why do you presume that they haven't? Every View has intrinsic full animation support, and of course there's a full simile of CoreAnimation in android.view.animation.
So why don't Android apps make use of it? Generally as a function of time, care, and craftsmanship. Too often Android apps were second tier, and many Android app writers were doing it to scratch an itch (versus for the iPhone where it is typically productized).
Android apps have the potential to be just as slick, just as polished, and just as nuanced and emotional as any iphone app. The state today is that they aren't, but that isn't any intrinsic failure of the platform.
Honestly, though, I find that sort of observation superficial and just incredibly boring. Every iphone user when they touch an Android device will immediately comment on the lack of slickness of the non-GPU using transitions. As a day to day Android user, using my phone for virtually everything and anything, this has absolutely ZERO impact on my enjoyment of the device. To say it is unnoticed is still giving it too much credit. It's superficial bedazzling that differentiates based upon the utterly irrelevant.
Animations are a core part of the design, they make it much easier for normal people to understand transitions and spatial interfaces in the UI, e.g. when going from one screen to another in a hierarchy. And they make it look nicer. (When was the last time a consumer bought a car by specs instead of by looks?)
First we had text-based UIs (hello /bin/sh). Then we had static GUIs. Now we have animated GUIs. Many old-school unix gurus couldn't see the need for GUIs, e.g. BrianK said "What You See Is All You Get" about WYSIWYG because he thought GUIs were more limited than text-based UIs. But for normal people, the 'G' in GUI was the important letter and is what enables them to use the computer.
edit: added missing word: 'limited'
Animations are a core part of the design
Don't stop with just technology, though. Even more powerful: culture and community.
Who is completely ignoring the design? My comment was specifically and only about essentially spurious, extraneous chrome -- literally interface bedazzling. It has nothing whatsoever to do with usability, and actually little to do with design.
You know I still marvel that Apple is held as the gold standard of design. The iPhones notification system is truly terrible. The messaging system is a god awful sin of design (for the longest time I simply could not believe that the screenshots posted were actually from the esteemed iPhone). Many of the transition effects stink of excess.
One person's excess is another's necessity. I haven't spent more than 10 minutes with an iPhone, so I'll comment on what I'm familiar with: Compiz. A lot of hackers call the wobbly windows effect in Compiz a waste of resources for a purely visual effect. Without wobbly windows, however, everything feels stiff to me. My workflow is more fluid with wobbly windows, making me more productive.
Comment #34 from 'Request for GPU support' thread linked in this comment thread by pkaler (http://code.google.com/p/android/issues/detail?id=6914)
> I agree with the common opinion here... It is very hard to pitch Android to someone who is not tech savy, simply because of the choppy and sluggish UI. This has also become the talk of the masses, most people who haven't even used Android say the following: "I've seen Android, its cool, but its not as smooth and stable as the iPhone." The iPhone's hardware fails a lot (I've exchanged my 3Gs twice in less than a year), it also freezes at times and drops calls, but what do people see and remember!? UI!! If android wishes to ever compete seriously with iOS, GPU acceleration is a must!
This is how most software purchase/development decisions are made, especially those with multi-million dollar consequences.
> Emotion is a huge factor when people choose what to buy — I’d say maybe even the biggest one. Apple understands this. All iOS devices — all Apple devices, for that matter — are designed with the emotional experience in mind. Why does almost everything in iOS animate? Why did Apple create CoreAnimation, and base UIKit app development so heavily upon it? Because animation, even in small unobtrusive doses, has an emotional affect. It results in a feeling.