Develop Reference

Why is it important that Java (and other JVM languages) is highly portable? [duplicate]

I've been thinking about it lately, and it seems to me that most advantages given to JIT compilation should more or less be attributed to the intermediate format instead, and that jitting in itself is not much of a good way to generate code.
So these are the main pro-JIT compilation arguments I usually hear:
Just-in-time compilation allows for greater portability. Isn't that attributable to the intermediate format? I mean, nothing keeps you from compiling your virtual bytecode into native bytecode once you've got it on your machine. Portability is an issue in the 'distribution' phase, not during the 'running' phase.
Okay, then what about generating code at runtime? Well, the same applies. Nothing keeps you from integrating a just-in-time compiler for a real just-in-time need into your native program.
But the runtime compiles it to native code just once anyways, and stores the resulting executable in some sort of cache somewhere on your hard drive. Yeah, sure. But it's optimized your program under time constraints, and it's not making it better from there on. See the next paragraph.
It's not like ahead-of-time compilation had no advantages either. Just-in-time compilation has time constraints: you can't keep the end user waiting forever while your program launches, so it has a tradeoff to do somewhere. Most of the time they just optimize less. A friend of mine had profiling evidence that inlining functions and unrolling loops "manually" (obfuscating source code in the process) had a positive impact on performance on his C# number-crunching program; doing the same on my side, with my C program filling the same task, yielded no positive results, and I believe this is due to the extensive transformations my compiler was allowed to make.
And yet we're surrounded by jitted programs. C# and Java are everywhere, Python scripts can compile to some sort of bytecode, and I'm sure a whole bunch of other programming languages do the same. There must be a good reason that I'm missing. So what makes just-in-time compilation so superior to ahead-of-time compilation?
EDIT To clear some confusion, maybe it would be important to state that I'm all for an intermediate representation of executables. This has a lot of advantages (and really, most arguments for just-in-time compilation are actually arguments for an intermediate representation). My question is about how they should be compiled to native code.
Most runtimes (or compilers for that matter) will prefer to either compile them just-in-time or ahead-of-time. As ahead-of-time compilation looks like a better alternative to me because the compiler has more time to perform optimizations, I'm wondering why Microsoft, Sun and all the others are going the other way around. I'm kind of dubious about profiling-related optimizations, as my experience with just-in-time compiled programs displayed poor basic optimizations.
I used an example with C code only because I needed an example of ahead-of-time compilation versus just-in-time compilation. The fact that C code wasn't emitted to an intermediate representation is irrelevant to the situation, as I just needed to show that ahead-of-time compilation can yield better immediate results.

Greater portability: The
deliverable (byte-code) stays
portable
At the same time, more platform-specific: Because the
JIT-compilation takes place on the
same system that the code runs, it
can be very, very fine-tuned for
that particular system. If you do
ahead-of-time compilation (and still
want to ship the same package to
everyone), you have to compromise.
Improvements in compiler technology can have an impact on
existing programs. A better C
compiler does not help you at all
with programs already deployed. A
better JIT-compiler will improve the
performance of existing programs.
The Java code you wrote ten years ago will run faster today.
Adapting to run-time metrics. A JIT-compiler can not only look at
the code and the target system, but
also at how the code is used. It can
instrument the running code, and
make decisions about how to optimize
according to, for example, what
values the method parameters usually
happen to have.
You are right that JIT adds to start-up cost, and so there is a time-constraint for it,
whereas ahead-of-time compilation can take all the time that it wants. This makes it
more appropriate for server-type applications, where start-up time is not so important
and a "warm-up phase" before the code gets really fast is acceptable.
I suppose it would be possible to store the result of a JIT compilation somewhere, so that it could be re-used the next time. That would give you "ahead-of-time" compilation for the second program run. Maybe the clever folks at Sun and Microsoft are of the opinion that a fresh JIT is already good enough and the extra complexity is not worth the trouble.

The ngen tool page spilled the beans (or at least provided a good comparison of native images versus JIT-compiled images). Executables that are compiled ahead-of-time typically have the following benefits:
Native images load faster because they don't have much startup activities, and require a static amount of fewer memory (the memory required by the JIT compiler);
Native images can share library code, while JIT-compiled images cannot.
Just-in-time compiled executables typically have the upper hand in these cases:
Native images are larger than their bytecode counterpart;
Native images must be regenerated whenever the original assembly or one of its dependencies is modified.
The need to regenerate an image that is ahead-of-time compiled every time one of its components is a huge disadvantage for native images. On the other hand, the fact that JIT-compiled images can't share library code can cause a serious memory hit. The operating system can load any native library at one physical location and share the immutable parts of it with every process that wants to use it, leading to significant memory savings, especially with system frameworks that virtually every program uses. (I imagine that this is somewhat offset by the fact that JIT-compiled programs only compile what they actually use.)
The general consideration of Microsoft on the matter is that large applications typically benefit from being compiled ahead-of-time, while small ones generally don't.

Simple logic tell us that compiling huge MS Office size program even from byte-codes will simply take too much time. You'll end up with huge starting time and that will scare anyone off your product. Sure, you can precompile during installation but this also has consequences.
Another reason is that not all parts of application will be used. JIT will compile only those parts that user care about, leaving potentially 80% of code untouched, saving time and memory.
And finally, JIT compilation can apply optimizations that normal compilators can't. Like inlining virtual methods or parts of the methods with trace trees. Which, in theory, can make them faster.

Better reflection support. This could be done in principle in an ahead-of-time compiled program, but it almost never seems to happen in practice.
Optimizations that can often only be figured out by observing the program dynamically. For example, inlining virtual functions, escape analysis to turn stack allocations into heap allocations, and lock coarsening.

Maybe it has to do with the modern approach to programming. You know, many years ago you would write your program on a sheet of paper, some other people would transform it into a stack of punched cards and feed into THE computer, and tomorrow morning you would get a crash dump on a roll of paper weighing half a pound. All that forced you to think a lot before writing the first line of code.
Those days are long gone. When using a scripting language such as PHP or JavaScript, you can test any change immediately. That's not the case with Java, though appservers give you hot deployment. So it is just very handy that Java programs can be compiled fast, as bytecode compilers are pretty straightforward.
But, there is no such thing as JIT-only languages. Ahead-of-time compilers have been available for Java for quite some time, and more recently Mono introduced it to CLR. In fact, MonoTouch is possible at all because of AOT compilation, as non-native apps are prohibited in Apple's app store.

I have been trying to understand this as well because I saw that Google is moving towards replacing their Dalvik Virtual Machine (essentially another Java Virtual Machine like HotSpot) with Android Run Time (ART), which is a AOT compiler, but Java usually uses HotSpot, which is a JIT compiler. Apparently, ARM is ~ 2x faster than Dalvik... so I thought to myself "why doesn't Java use AOT as well?".
Anyways, from what I can gather, the main difference is that JIT uses adaptive optimization during run time, which (for example) allows ONLY those parts of the bytecode that are being executed frequently to be compiled into native code; whereas AOT compiles the entire source code into native code, and code of a lesser amount runs faster than code of a greater amount.
I have to imagine that most Android apps are composed of a small amount of code, so on average it makes more sense to compile the entire source code to native code AOT and avoid the overhead associated from interpretation / optimization.

It seems that this idea has been implemented in Dart language:
https://hackernoon.com/why-flutter-uses-dart-dd635a054ebf
JIT compilation is used during development, using a compiler that is especially fast. Then, when an app is ready for release, it is compiled AOT. Consequently, with the help of advanced tooling and compilers, Dart can deliver the best of both worlds: extremely fast development cycles, and fast execution and startup times.

One advantage of JIT which I don't see listed here is the ability to inline/optimize across separate assemblies/dlls/jars (for simplicity I'm just going to use "assemblies" from here on out).
If your application references assemblies which might change after install (e. g. pre-installed libraries, framework libraries, plugins), then a "compile-on-install" model must refrain from inlining methods across assembly boundaries. Otherwise, when the referenced assembly is updated we would have to find all such inlined bits of code in referencing assemblies on the system and replace them with the updated code.
In a JIT model, we can freely inline across assemblies because we only care about generating valid machine code for a single run during which the underlying code isn't changing.

The difference between platform-browser-dynamic and platform-browser is the way your angular app will be compiled.
Using the dynamic platform makes angular sending the Just-in-Time compiler to the front-end as well as your application. Which means your application is being compiled on client-side.
On the other hand, using platform-browser leads to an Ahead-of-Time pre-compiled version of your application being sent to the browser. Which usually means a significantly smaller package being sent to the browser.
The angular2-documentation for bootstrapping at https://angular.io/docs/ts/latest/guide/ngmodule.html#!#bootstrap explains it in more detail.

What is the purpose of bytecode in Java?

Given that I can compile 300 classes in seconds, an implementation of Java could just be given Java source files instead of bytecode as an input, then compile and cache the input source code, and never compile it again (e.g python does this, and lots of language implementations do the same except don't even bother to cache):
This initial compilation experience would be equivalent to the installation process that users are already used to
This would remove the need for implementing the non trivial task of verification in the bytecode interpreter (which really is just reimplementing parts of the compile time checks), reducing implementation complexity.
Java as it currently is, verifies the input bytecode every time it starts, even if it already verified it before. Point 2 would of course reduce startup times because it eliminates this step (although the current Java platform could also cache the "checked" status somewhere to reduce startup times, I'm not sure if it does this)
This would allow implementations to compile however they want (or not at all), e.g for performance. Android doesn't even use Java bytecode, it uses dalvik bytecode, because they claim it's more suitable for their needs (e.g more performant on their hardware). If bytecode didn't exist, this design decision made by Google would have been completely transparent.
This would promote open source
This answers why distribute bytecode instead of native code, but to be clear, I'm wondering why even have a compiled format for distribution at all? Assuming that compilation is important, why not just have the runtime compile source and cache it?
The only remaining rationale I can come up with is for obfuscation, but...
the way current compilers compile, the code can be mechanically decompiled pretty accurately
source code can be obfuscated too
...so this point is reduced to that intuition would say that bytecode is more complicated than source code, thus having a bytecode distribution format allows to trick businessmen into thinking their IP is protected (i.e bytecode would be to "add value", but there is no technical reason for it).
Why is the Java platform designed for distributing bytecode to the user, as opposed to distributing source code to the user? I could not find an explanation of this anywhere on the internet. Is there a big reason I am missing here?
If you give a reason, you should probably state whether it's a reason that the language designers initially had, or a reason that is still valid today.

You are thinking just inside your little world. There are some compelling reasons to compile the source and deliver bytecode instead:
Download time (Applets were supposed to become a widely accepted web technology) - the user has no need for the source, so why retain the source? Reducing the amount of information transmitted means faster downloads.
Reduces startup time. Compiling on every run, takes additional time. If you can compile 300 classes a second that would mean an additional 5-10 Seconds startup time on the JRE alone nowadays. And machines were a little slower in 1995, you know.
Java is aimed at a multitude of platforms. Some platforms are not as powerful as your PC. Think of embedded and mobile devices. They may have neither the storage nor the power to compile code.
Bytecode allows any language to be compiled to bytecode - not just Java. There are plenty of other languages that can compile to bytecode. Would you prefer to install a new compiler for each of them?
Companies are often reluctant to give "the source" out of their hands. Java would have more acceptance problems if programs were to be delivered "in source".
Bytecode is a form of machine code simple enough to execute in hardware directly (There are a few embedded designs that have partial native bytecode support).
I'm sure there are more pros for bytecode I haven't even though about.

How does C differ from Java's compilation model?

What the differences between classical compilation model (C, C++, etc.) and the Java compilation model?

A proper answer to your question could take several hundred pages to answer, but I'll try to sum it up in a few paragraphs.
Basically, the "classic compilation model" you refer to takes as input human-written source code and emits machine code, which can be loaded and run without further translation of the machine code. One ramification of this is that the resulting machine code can only be run on compatible hardware and can only be run within a compatible operating system.
The Java compilation model takes human-written source code as input and emits not machine code, but so-called "byte code". Byte code cannot be directly executed on a machine. Instead, it needs to be translated once again by another compiler to machine code, or interpreted on-the-fly by a device that executes instructions on the machine that correspond to the instructions in the byte code. The latter device is often referred to as a Virtual Machine. One ramification of this model is that the byte code can be "run" on any platform that has either a byte code compiler or virtual machine written for it. This gives Java the appearance and effect of complete portability, where there is no such portability implied by the machine code emitted by a C++ compiler stack.

Two aspects play into the C (and C++) compilation model. One is its longer history than Java, meaning that it caters to very low-powered compilers and machines. The second is the compilation target, which is usually low-level machine code.
To target low-memory compiler environments, C code must be readable from top to bottom, with no backtracking. This means that you have to follow a strict discipline for the order of declarations. (C++ relaxes this a little bit for class definitions.) Further more, each source file must be compilable as an independent translation unit which need not know anything about other source files.
Second, because C targets low-level machine code, this means that each translation unit contains essentially no metadata, in stark contrast to Java class files. This necessitates a stronger coding discipline in which each translation unit must be provided with the necessary declarations. The compiler cannot just scan all the other files in order to get the required information; it is up to the user to supply it. (C++ enforces this more rigidly, in C you can get away with nasty errors by forgetting a declaration.)
Bear in mind that a C program has to be fully compiled and linked at compile time, so a lot of information has to be available already at that point. Java programs can load classes at runtime, and Java execution generally performs more "fitting" operations (casting, essentially, as opposed to static linking in C) at runtime. The more sophisticated runtime environment of Java allows for a more flexible and modular compilation model.

I am going to be brave and compare performance. ;)
The Java compiler javac does little optimisation preferring to syntax check code. It does all the reasonable checks required to ensure it will run on a JVM, and some constant evaluation and that's about it.
Most of the smart compilation is done by the JIT which can perform dynamic complication based on how the program is used. This allows it to inline "virtual" methods, for example, even if the caller and callee are in different libraries.
The C/C++ compiler performs significant static analysis up front. This means a program will run at almost full speed right from the start. The CPU performs some dynamic optimisation with instruction re-ordering and branch prediction. While C/C++ lacks dynamic optimisation, it gains from by making low level access to the system much easier. (Its usually not impossible in Java, but low level operations which are trivial in C/C++ can be complex and obscure in Java) It also provides more ways to do the same thing allowing you to choose the optimal solution to your problem.
When Java is likely to be faster.
If your style of programming suits Java and you only use the sort of features Java supports, Java is likely to be marginally faster (due to dynamic compilation) i.e. you wouldn't use C/C++ to their full potential anyway.
If your code contains lots of dead code (possibly only known to be dead at run time) Java does a good job at eliminating this. (IMHO A high percentage of micro-benchmarks which suggest Java is faster than C++ are of this type)
You have a very limited time and/or resources to implement your application. (In which case an even higher level language might be better) i.e. You don't have time to optimise your code much and you need to write safe abstracted code.
When C/C++ is likely to be faster.
If you use most of the functionality C/C++ provides. Something more advanced programmers tend to do.
If startup time matters.
If you need to be creative about algorithms or data structures.
If you can exploit a low level hardware feature, like direct access to devices.

For short, "classical" compilation (which is a temp term provided by the material because they don't have a real word for it), is basically compiling against a real device (in our case a machine with a physical processor). Instead, Java compiles to code against a virtual device, which is software installed on a machine. The virtual device is what changes and targets the real machine.
In this way your hardware is abstracted. This is why Java can work on "any" machine.

Basically, there are two kinds of magic. Machine magic is only understood by certain wizards. JVM Bytecode magic is understood by a special kind of wizard that you have to hire in order to make the machine wizard able to cast spells that make your computer do things. C and C++ compilers generally emit the machine kind, whereas Java compilers emit JVM Bytecode.

C/C++ gets compiled before execution.
Java gets compiled while executing.
Of course, neither language mandates a certain way of being compiled.

There is no difference. Both convert source code that a human understands, to a machine code that some machine understands. In Java's case it targets a virtual machine, i.e. a program instead of a piece of silicon.
Of course there's nothing to prevent a piece of silicon from understanding JVM byte code (in which case you could rename it from 'byte code' to 'machine code'). And conversely, there's nothing to prevent a compiler from converting C/C++ code to JVM byte code.
Both have a runtime and both require you to tell it which parts of the runtime you intend to use.
I really think you intended to ask a different question.

Compiled vs. Interpreted Languages

I'm trying to get a better understanding of the difference. I've found a lot of explanations online, but they tend towards the abstract differences rather than the practical implications.
Most of my programming experiences has been with CPython (dynamic, interpreted), and Java (static, compiled). However, I understand that there are other kinds of interpreted and compiled languages. Aside from the fact that executable files can be distributed from programs written in compiled languages, are there any advantages/disadvantages to each type? Oftentimes, I hear people arguing that interpreted languages can be used interactively, but I believe that compiled languages can have interactive implementations as well, correct?

A compiled language is one where the program, once compiled, is expressed in the instructions of the target machine. For example, an addition "+" operation in your source code could be translated directly to the "ADD" instruction in machine code.
An interpreted language is one where the instructions are not directly executed by the target machine, but instead read and executed by some other program (which normally is written in the language of the native machine). For example, the same "+" operation would be recognised by the interpreter at run time, which would then call its own "add(a,b)" function with the appropriate arguments, which would then execute the machine code "ADD" instruction.
You can do anything that you can do in an interpreted language in a compiled language and vice-versa - they are both Turing complete. Both however have advantages and disadvantages for implementation and use.
I'm going to completely generalise (purists forgive me!) but, roughly, here are the advantages of compiled languages:
Faster performance by directly using the native code of the target machine
Opportunity to apply quite powerful optimisations during the compile stage
And here are the advantages of interpreted languages:
Easier to implement (writing good compilers is very hard!!)
No need to run a compilation stage: can execute code directly "on the fly"
Can be more convenient for dynamic languages
Note that modern techniques such as bytecode compilation add some extra complexity - what happens here is that the compiler targets a "virtual machine" which is not the same as the underlying hardware. These virtual machine instructions can then be compiled again at a later stage to get native code (e.g. as done by the Java JVM JIT compiler).

A language itself is neither compiled nor interpreted, only a specific implementation of a language is. Java is a perfect example. There is a bytecode-based platform (the JVM), a native compiler (gcj) and an interpeter for a superset of Java (bsh). So what is Java now? Bytecode-compiled, native-compiled or interpreted?
Other languages, which are compiled as well as interpreted, are Scala, Haskell or Ocaml. Each of these languages has an interactive interpreter, as well as a compiler to byte-code or native machine code.
So generally categorizing languages by "compiled" and "interpreted" doesn't make much sense.

Start thinking in terms of a: blast from the past
Once upon a time, long long ago, there lived in the land of computing
interpreters and compilers. All kinds of fuss ensued over the merits of
one over the other. The general opinion at that time was something along the lines of:
Interpreter: Fast to develop (edit and run). Slow to execute because each statement had to be interpreted into
machine code every time it was executed (think of what this meant for a loop executed thousands of times).
Compiler: Slow to develop (edit, compile, link and run. The compile/link steps could take serious time). Fast
to execute. The whole program was already in native machine code.
A one or two order of magnitude difference in the runtime
performance existed between an interpreted program and a compiled program. Other distinguishing
points, run-time mutability of the code for example, were also of some interest but the major
distinction revolved around the run-time performance issues.
Today the landscape has evolved to such an extent that the compiled/interpreted distinction is
pretty much irrelevant. Many
compiled languages call upon run-time services that are not
completely machine code based. Also, most interpreted languages are "compiled" into byte-code
before execution. Byte-code interpreters can be very efficient and rival some compiler generated
code from an execution speed point of view.
The classic difference is that compilers generated native machine code, interpreters read source code and
generated machine code on the fly using some sort of run-time system.
Today there are very few classic interpreters left - almost all of them
compile into byte-code (or some other semi-compiled state) which then runs on a virtual "machine".

The extreme and simple cases:
A compiler will produce a binary executable in the target machine's native executable format. This binary file contains all required resources except for system libraries; it's ready to run with no further preparation and processing and it runs like lightning because the code is the native code for the CPU on the target machine.
An interpreter will present the user with a prompt in a loop where he can enter statements or code, and upon hitting RUN or the equivalent the interpreter will examine, scan, parse and interpretatively execute each line until the program runs to a stopping point or an error. Because each line is treated on its own and the interpreter doesn't "learn" anything from having seen the line before, the effort of converting human-readable language to machine instructions is incurred every time for every line, so it's dog slow. On the bright side, the user can inspect and otherwise interact with his program in all kinds of ways: Changing variables, changing code, running in trace or debug modes... whatever.
With those out of the way, let me explain that life ain't so simple any more. For instance,
Many interpreters will pre-compile the code they're given so the translation step doesn't have to be repeated again and again.
Some compilers compile not to CPU-specific machine instructions but to bytecode, a kind of artificial machine code for a ficticious machine. This makes the compiled program a bit more portable, but requires a bytecode interpreter on every target system.
The bytecode interpreters (I'm looking at Java here) recently tend to re-compile the bytecode they get for the CPU of the target section just before execution (called JIT). To save time, this is often only done for code that runs often (hotspots).
Some systems that look and act like interpreters (Clojure, for instance) compile any code they get, immediately, but allow interactive access to the program's environment. That's basically the convenience of interpreters with the speed of binary compilation.
Some compilers don't really compile, they just pre-digest and compress code. I heard a while back that's how Perl works. So sometimes the compiler is just doing a bit of the work and most of it is still interpretation.
In the end, these days, interpreting vs. compiling is a trade-off, with time spent (once) compiling often being rewarded by better runtime performance, but an interpretative environment giving more opportunities for interaction. Compiling vs. interpreting is mostly a matter of how the work of "understanding" the program is divided up between different processes, and the line is a bit blurry these days as languages and products try to offer the best of both worlds.

From http://www.quora.com/What-is-the-difference-between-compiled-and-interpreted-programming-languages
There is no difference, because “compiled programming language” and
“interpreted programming language” aren’t meaningful concepts. Any
programming language, and I really mean any, can be interpreted or
compiled. Thus, interpretation and compilation are implementation
techniques, not attributes of languages.
Interpretation is a technique whereby another program, the
interpreter, performs operations on behalf of the program being
interpreted in order to run it. If you can imagine reading a program
and doing what it says to do step-by-step, say on a piece of scratch
paper, that’s just what an interpreter does as well. A common reason
to interpret a program is that interpreters are relatively easy to
write. Another reason is that an interpreter can monitor what a
program tries to do as it runs, to enforce a policy, say, for
security.
Compilation is a technique whereby a program written in one language
(the “source language”) is translated into a program in another
language (the “object language”), which hopefully means the same thing
as the original program. While doing the translation, it is common for
the compiler to also try to transform the program in ways that will
make the object program faster (without changing its meaning!). A
common reason to compile a program is that there’s some good way to
run programs in the object language quickly and without the overhead
of interpreting the source language along the way.
You may have guessed, based on the above definitions, that these two
implementation techniques are not mutually exclusive, and may even be
complementary. Traditionally, the object language of a compiler was
machine code or something similar, which refers to any number of
programming languages understood by particular computer CPUs. The
machine code would then run “on the metal” (though one might see, if
one looks closely enough, that the “metal” works a lot like an
interpreter). Today, however, it’s very common to use a compiler to
generate object code that is meant to be interpreted—for example, this
is how Java used to (and sometimes still does) work. There are
compilers that translate other languages to JavaScript, which is then
often run in a web browser, which might interpret the JavaScript, or
compile it a virtual machine or native code. We also have interpreters
for machine code, which can be used to emulate one kind of hardware on
another. Or, one might use a compiler to generate object code that is
then the source code for another compiler, which might even compile
code in memory just in time for it to run, which in turn . . . you get
the idea. There are many ways to combine these concepts.

The biggest advantage of interpreted source code over compiled source code is PORTABILITY.
If your source code is compiled, you need to compile a different executable for each type of processor and/or platform that you want your program to run on (e.g. one for Windows x86, one for Windows x64, one for Linux x64, and so on). Furthermore, unless your code is completely standards compliant and does not use any platform-specific functions/libraries, you will actually need to write and maintain multiple code bases!
If your source code is interpreted, you only need to write it once and it can be interpreted and executed by an appropriate interpreter on any platform! It's portable! Note that an interpreter itself is an executable program that is written and compiled for a specific platform.
An advantage of compiled code is that it hides the source code from the end user (which might be intellectual property) because instead of deploying the original human-readable source code, you deploy an obscure binary executable file.

A compiler and an interpreter do the same job: translating a programming language to another pgoramming language, usually closer to the hardware, often direct executable machine code.
Traditionally, "compiled" means that this translation happens all in one go, is done by a developer, and the resulting executable is distributed to users. Pure example: C++.
Compilation usually takes pretty long and tries to do lots of expensive optmization so that the resulting executable runs faster. End users don't have the tools and knowledge to compile stuff themselves, and the executable often has to run on a variety of hardware, so you can't do many hardware-specific optimizations. During development, the separate compilation step means a longer feedback cycle.
Traditionally, "interpreted" means that the translation happens "on the fly", when the user wants to run the program. Pure example: vanilla PHP. A naive interpreter has to parse and translate every piece of code every time it runs, which makes it very slow. It can't do complex, costly optimizations because they'd take longer than the time saved in execution. But it can fully use the capabilities of the hardware it runs on. The lack of a separrate compilation step reduces feedback time during development.
But nowadays "compiled vs. interpreted" is not a black-or-white issue, there are shades in between. Naive, simple interpreters are pretty much extinct. Many languages use a two-step process where the high-level code is translated to a platform-independant bytecode (which is much faster to interpret). Then you have "just in time compilers" which compile code at most once per program run, sometimes cache results, and even intelligently decide to interpret code that's run rarely, and do powerful optimizations for code that runs a lot. During development, debuggers are capable of switching code inside a running program even for traditionally compiled languages.

First, a clarification, Java is not fully static-compiled and linked in the way C++. It is compiled into bytecode, which is then interpreted by a JVM. The JVM can go and do just-in-time compilation to the native machine language, but doesn't have to do it.
More to the point: I think interactivity is the main practical difference. Since everything is interpreted, you can take a small excerpt of code, parse and run it against the current state of the environment. Thus, if you had already executed code that initialized a variable, you would have access to that variable, etc. It really lends itself way to things like the functional style.
Interpretation, however, costs a lot, especially when you have a large system with a lot of references and context. By definition, it is wasteful because identical code may have to be interpreted and optimized twice (although most runtimes have some caching and optimizations for that). Still, you pay a runtime cost and often need a runtime environment. You are also less likely to see complex interprocedural optimizations because at present their performance is not sufficiently interactive.
Therefore, for large systems that are not going to change much, and for certain languages, it makes more sense to precompile and prelink everything, do all the optimizations that you can do. This ends up with a very lean runtime that is already optimized for the target machine.
As for generating executables, that has little to do with it, IMHO. You can often create an executable from a compiled language. But you can also create an executable from an interpreted language, except that the interpreter and runtime is already packaged in the exectuable and hidden from you. This means that you generally still pay the runtime costs (although I am sure that for some language there are ways to translate everything to a tree executable).
I disagree that all languages could be made interactive. Certain languages, like C, are so tied to the machine and the entire link structure that I'm not sure you can build a meaningful fully-fledged interactive version

This is one of the biggest misunderstood things in computer science as I guess.
Because interpretation and compilation are completely two different things, which we can't compare in this way.
The compilation is the process of translating one language into another language. There are few types of compilations.
Compiling - Translate high-level language into machine/byte code (ex: C/C++/Java)
Transpiling - Translate high-level language into another high-level language (ex: TypeScript)
Interpretation is the process of actually executing the program. This may happen in few different ways.
Machine level interpretation - This interpretation happens to the code which is compiled into machine code. Instructions are directly interpreted by the processor. Programming languages like C/C++ generate machine code, which is executable by the processor. So the processor can directly execute these instructions.
Virtual machine level interpretation - This interpretation happens to the code which is not compiled into the machine level (processor support) code, but into some intermediate-level code. This execution is done by another software, which is executed by the processor. At this time actually processor doesn't see our application. It just executing the virtual machine, which is executing our application. Programming languages like Java, Python, C# generate a byte code, which is executable by the virtual interpreter/machine.
So at the end of the day what we have to understand is, all the programming languages in the world should be interpreted at some time. It may be done by a processor(hardware) or a virtual machine.
The compilation is just the process of bringing the high-level code we write that is human-understandable into some hardware/software machine-understandable level.
These are completely two different things, which we can't compare. But that terminology is pretty much good to teach beginners how programming languages work.
PS:
Some programming languages like Java have a hybrid approach to do this. First, compile the high-level code into byte code which is virtual-machine readable. And on the fly, a component called the JIT compiler compiles byte-code into machine code. Specifically, code lines that are executed again and again many times are get translated into the machine language, which makes the interpretation process much faster. Because hardware processor is always much faster than virtual interpreter/processor.
How Java JIT compiler works

It's rather difficult to give a practical answer because the difference is about the language definition itself. It's possible to build an interpreter for every compiled language, but it's not possible to build an compiler for every interpreted language. It's very much about the formal definition of a language. So that theoretical informatics stuff noboby likes at university.

The Python Book © 2015 Imagine Publishing Ltd, simply distunguishes the difference by the following hint mentioned in page 10 as:
An interpreted language such as Python is one where the source code is converted to machine code and then executed each time the program runs. This is different from a compiled language such as C, where the source code is only converted to machine code once – the resulting machine code is then executed each time the program runs.

Compile is the process of creating an executable program from code written in a compiled programming language. Compiling allows the computer to run and understand the program without the need of the programming software used to create it. When a program is compiled it is often compiled for a specific platform (e.g. IBM platform) that works with IBM compatible computers, but not other platforms (e.g. Apple platform).
The first compiler was developed by Grace Hopper while working on the Harvard Mark I computer. Today, most high-level languages will include their own compiler or have toolkits available that can be used to compile the program. A good example of a compiler used with Java is Eclipse and an example of a compiler used with C and C++ is the gcc command. Depending on how big the program is it should take a few seconds or minutes to compile and if no errors are encountered while being compiled an executable file is created.check this information

Short (un-precise) definition:
Compiled language: Entire program is translated to machine code at once, then the machine code is run by the CPU.
Interpreted language: Program is read line-by-line and as soon as a line is read the machine instructions for that line are executed by the CPU.
But really, few languages these days are purely compiled or purely interpreted, it often is a mix. For a more detailed description with pictures, see this thread:
What is the difference between compilation and interpretation?
Or my later blog post:
https://orangejuiceliberationfront.com/the-difference-between-compiler-and-interpreter/

What can you not do on the Dalvik VM (Android's VM) that you can in Sun VM?

I know that you can run almost all Java in Dalvik's VM that you can in Java's VM but the limitations are not very clear. Has anyone run into any major stumbling blocks? Any major libraries having trouble? Any languages that compile to Java byte code (Scala, Jython etc...) not work as expected?

There is a number of things that Dalvik will not handle or will not handle quite the same way as standard Java bytecode, though most of them are quite advanced.
The most severe example is runtime bytecode generation and custom class loading. Let's say you would like to create some bytecode and then use classloader to load it for you, if that trick works on your normal machine, it is guaranteed to not work on Dalvik, unless you change your bytecode generation.
That prevents you from using certain dependency injection frameworks, most known example being Google Guice (though I am sure some people work on that). On the other hand AspectJ should work as it uses bytecode instrumentation as a compilation step (though I don't know if anyone tried).
As to other jvm languages -- anything that in the end compiles to standard bytecode and does not use bytecode instrumentation at runtime can be converted to Dalvik and should work. I know people did run Jython on Android and it worked ok.
Other thing to be aware of is that there is no just in time compilation. This is not strictly Dalviks problem (you can always compile any bytecode on the fly if you wish) but that Android does not support that and is unlikely to do so. In the effect while microbenchmarking for standard Java was useless -- components had different runtime characterstics in tests than as parts of larger systems -- microbenchmarks for Android phones totally make sense.

If you see "Dalvik Virtual Machine internals" Google IO session, you can find Dalvik does not support generational GC.
So, it could degrade performance of frequent object creation and deletion. Java VM supports generational GC so, it would show better GC performance for the same situation.
And also, Dalvik uses trace-granuality JIT instead of method granuality JIT.

Another thing that I guess could be added here is that Dalvik apparently does not preserve field order when listing the fields of a class using the reflection API. Now, the reflection API does not make any guarantees on it anyway (so ideally you shouldn't depend on it anyway), but most of the other VMs out there do preserve the order.

Just to add to the conversation, not intended to revive an old thread. I just ran across this in my search, and want to add that Jython does not work out of the box with Dalvik either. Simply trying to do a hello world example will yield the following: