Develop Reference

Is there a thing such as "packaged web app"?

Let's say I have a HTML/JS application, which is on a website and runs without any server-side component (except the web server, of course). So it's basically just a bunch of .html, .css and .js files (and some others, like images or SQLite databases)
Someday, an user comes to me to ask me if I can make an offline version, so they can access the app when they are offline.
My first reaction would be to provide them all the files of the app in a .zip file or something, but this is not really good: it's not "clear" to an average user, and the files and code are apparent (which could be "dangerous" if the user edits or deletes them by accident).
So I was wondering: is there a kind of format that would makes me able to store the files of the app in a packaged file?
The file would be something like a .zip file, but when you click to the file, instead of opening up the archive, it opens the /index.html (or similar) which is inside.
So for the user, it would be similar to open an .exe or a single .html file, but still that's not a native app, as that's still a browser app.
Given that I know about:
Apache Cordova; but most of the time, I just use the browser functions, and I don't have a need to use advanced native functionalities;
UWP and PWA, but I don't know if it's exactly what I want (I imagine it as a mix of website and mobile app, but I'm surely wrong);
Open Web App (a standard from Mozilla), but based on what I've read about it, it seems deprecated and not used anymore. Plus, I don't really look forward to "certify" my app... I just want to package some files.

PWA I don't think is a solution to your problem as you wont be able to redistribute the code as an executable package.
UWP would be too Windows specific and might make you lose the platform independence that you already have since you are using plain HTML and JS, plus from what I understand you would need to distribute the app on Windows Store.
I believe packaging as an Electron app would be the best solution as you can provide executable files that can be executed in almost all platforms.

Electron would be the best solution for you. I had a client that first wanted to run app on php server and eventually they figured out they need it as a desktop app. It was easy to rewrite php stuff to node and package as Electron app. You have access to everything files, system etc. so you can do more if needed.
Another solution would be to just inline everything into a single html file.

Android and Web App sharing Kotlin code

One of the major “selling” points of Kotlin appears to be its ability to compile both to JVM or Android and to JavaScript for the web. To make use of this, it should be possible to have a code base where some files are shared between an Android App and a browser Web App.
However, so far I found little details on how such a thing would be set up, in particularly when working with Android Studio and its underlying Gradle setup, starting from a a run of the Android Studio New Project Wizard. I don't mind if I can only build the Web App on the command line, but I'd like to maintain the Android debugging hookups that Android Studio provides. So far I know very little about Gradle and typical idioms for its use.
I'm sure that I'm not the first person to have this idea, so I'd like to know some best practices on how to set this up. Questions that come to my mind include the following:
Do I mix the kotlin2js and the kotlin-android plugin in a single build file, or do I need to have multiple build files (perhaps I should say “modules” or “projects” except I don't know which)?
If I have multiple build files, should that be two (one Android one Web) or three (one more for shared things)?
If it is two build files, how do I reference the shared sources?
If it is three build files, which plugin(s) do I use for the shared one?
Do I need to split my sources over three different source trees? Are there any conventions how these should be called?
Do I need to split my classes into three groups of packages, or can code for different targets coexist in the same package?
What configuration settings do I need to tweak to make the IDE aware of the layout of my project?
I've read the following relevant documentation, among other:
https://kotlinlang.org/docs/reference/using-gradle.html
https://docs.gradle.org/current/userguide/intro_multi_project_builds.html
https://docs.gradle.org/current/userguide/composite_builds.html
https://kotlinlang.org/docs/tutorials/kotlin-android.html

I would recommend using IDEA wizard to create a simple multiplatfrom project for you (File -> New -> Project -> Kotlin -> Kotlin (Multiplatform - experimental) ). Community edition should suffice.
Answering your questions:
You don't mix plugins. You create a separate module for your common code and use 'kotlin-platform-common' plugin for it.
Three modules, special plugin 'kotlin-platform-common'
Use common sense for source splitting. Put whatever you want/able to reuse in the common code. Put platform-specific code in platform modules.
No package restrictions. You can put everything in the same package if you so desire =)
Pretty sure it should just work. If not, try re-importing.

Is there a way to read a local netcdf file from a web application?

We are building a web service for data analysis and would like to access netcdf files from the local machine where the browser is running. Javascript offers a file browser, but (for security reasons as I learned) it will automatically upload a file after selection, instead of allowing (read-only) access to it. This presents a show-stopper, because the netcdf files can be HUGE. Note, that the netcdf format and API explicitly allow slicing and extraction of individual variables, which is one reason why the format is so popular.
Now, some research into this issue revealed that the server-client architecture normally doesn't allow access to the local file structure to prevent spying. On the other hand, in HTML5 there is a file API which supports exactly the kind of operations we need -- except that you can access portions of a file by specifying byte ranges, but there is no netcdf API available; hence one would be left again with copying the entire file before being able to slice it on the server.
Of course, the other option are web services such as OpenDAP which are meant to do exactly what we want, i.e. access parts of a netcdf file over the internet. However, this of course requires that every user would have to install the OpenDAP server before they could access their local files from within the web service. (Or, at a minimum they would have to install a web server so that one could access the file via http://localhost/...).
So: does anyone know of a solution to read specific portions of a local netcdf file from a web application? Specifically, are there javascript tools available for this?

Currently you can use netcdfjs, that it's a javascript library that allows to read NetCDF v3 files. Because it's a NodeJS package you can run it server side, and for regular size files you can read it online, here you have an example.

There are no Javascript tools to read NetCDF, but there are Java libraries: http://www.unidata.ucar.edu/downloads/netcdf/index.jsp.
You could deploy a Java Applet to read the NetCDF file locally. From there you could have the Applet try to process the file, or communicate with a backend service, or even try to call on the Applet from Javascript: http://docs.oracle.com/javase/tutorial/deployment/applet/invokingAppletMethodsFromJavaScript.html.

There are no specific Javascript libs for reading NetCDF, as far as I know. But let me drop some thoughts:
As you say, you can install a local webserver, and use a server-side language to execute a program that reads the file in your disk. You could use a command-line software called ncdump-json. I intentionally wrote this software for that purpose.
You say that you don't want to install software like OpenDAP or webservers, but maybe a desktop app, a standalone .exe would be OK for you. Using node.js plus chromium (this is called nw.js) you have a program that is at the same time a nodejs server, and a chromium browser. This way you can write a web application that reads the data (e.g. using the embedded nodejs server and spawning ncdump-json...).
And now the simplest scenario, a 100% client-side solution... I think this can only be achieved using Javascript... or anything that can "compile to" Javascript. We need to have the netCDF libraries in Javascript, is that possible? I guess so. Emscripten is a LLVM-to-Javascript compiler, and netCDF is C so it has to be possible to compile it with clang/LLVM, therefore it has to be possible to use Emscripten to have a netCDF JS library version with almost no effort, without having to write from scratch (and also maintain) a JS port. If I'm not wrong, to random access the files there's the method slice in the html5 file api, so that should not be a problem.
Hope I helped.

How to include external Javascript in GWT offline application

I have a GWT (well, GXT) application that uses an external JavaScript library to add functionality to my app. My application must work offline, too, and herein lies my problem.
I am aware that adding files to the public folder will make them accessible by my GWT app, but this will not work in case of offline use. GWT compiles my app to make it available offline without problem, but it doesn't include the external JavaScript library.
So, whenever I work within the application and reach the point where said library is needed, the browser will attempt a GET request because the library hasn't been loaded yet and doesn't remain in the cache of the browser reliably.
Is there a way to add the library to my app so that it will be cached together with my GWT app? The library consists of several folders, JS files, images, CSS, etc. My only idea is to dynamically create an Appcache Manifest that dumps ALL files in the browser cache.. in which case I'm scared of breaking the GWT offline functionality.

Yes you can generate a manifest at compile time. Just use a linker that extends com.google.gwt.core.ext.linker.AbstractLinker.
See for example this example manifest linker
or see Writing a GWT Linker
or see this stackoverflow thread
I do that to include google fonts and to produce a manifest that will only include files for that specific language permutation.

Javascript for command line utilities

Given a need to write command line utilities to do common tasks like uploading files to a remote FTP site, downloading data from a remote MySQL database etc.
Is it practical to use JavaScript for this sort of thing? I know there are JavaScript interpreters that can be run from the command line, but are there libraries for things like FTP and database access the way there are for e.g. Java? If so, what's the best place to look for them? (Google searches with JavaScript in the keywords always seem to return many pages of browser specific things.)
And is there a way to package a JavaScript program up as a standalone executable on Windows?
Update: I've decided Python is a better tool for this kind of job, but the answers to the original question are still good ones.

Standalone executable?
By the way you ask the question, I'm not sure if you are aware, but the Windows Script Host - included in Windows - allows you to run .js files from the command-line. Your javascript will not be an executable, it will remain a script, a text file. The script runs within cscript.exe, which is provided by WSH. There's no compilation required. Maybe you knew all that.
I use Javascript this way for various utilities on Windows.
I think your instinct is right on the availability of libraries. You are sort of on your own to find all those things. Although, once you find them, it's not hard to package Javascript libraries as COM components and allow re-use from anywhere. See here for an example of packaging the Google Diff/Patch/Match Javascript library in COM.
Addendum: Once a bit of code is available within COM, it can be consumed by any Javascript running on the machine. Some examples of COM objects available to Javascript scripts running in WSH:
MSXML2.XMLHTTP object - used in AJAX, but can be used for any HTTP communication. There also an object for the XSLT engine so you can do transforms from script.
Excel.Application - allows you to open up Excel spreadsheets and automate them from Javascript.
Communicator.UIAutomation - automate MS Communicator (send IM's via script)
COM objects for Google Earth.
SlowAES - an all-Javascript implementation of AES encryption.

You can use Rhino to compile Javascript into Java byte code, and get access to all Java libraries.
Or you could use JScript.net, and get access to the .net libraries. .net includes a jsc.exe that produces exe-files.
Both of these requires the respective framework to be installed to be able to run.

Node.js is by far the best environment for running non-browser JS. I've used Rhino and SpiderMonkey, and there's a pretty huge difference in everything from the basics like how errors are handled to the size of the community using the tool. Node is pitched for "server-side" JS - building server apps in JS. It's great for this. But it works equally well for building command line tools.
The NPM package manager (bundled with Node) provides a nice global directory for finding and installing packages. It works much better than other language equivalents like PECL / Pear / CPAN / etc. Several high quality tools like JSHint, The Jade templating language, and the CoffeeScript compiler are all already available through NPM/Node:
npm install -g jshint, coffee-script, jade
jshint my_code.js
jade < my.jade > my.html
For args parsing, there are packages like commander.js. I currently use a heavily extended version of Commander in my underscore-cli command-line tool.
For messing with JSON or for doing command-line JS work (similar to "perl -pe"), check out underscore-cli - It's a really powerful tool for processing JSON data, processing underscore templates, and running JS expressions from the command-line. I use it for 1001 different things that would otherwise be really annoying to accomplish.

Rhino is bundled with JDK 1.6, jrunscript.exe in the bin directory will allow you to run any Javascript you want. Since it runs under Java you get access to any Java libraries that you may have.
We use it from the command line extensively. It's very good at that.

One way is to write these utilities as AIR applications - They can be written in JavaScript and need not have a UI. They have access to the command line, and there are existing ActionScript 3 libraries that can handle FTP etc. These ActionScript APIs can be called from JS, in AIR applications. AIR applications also have access to a sqlite database.

jslibs is a good standalone JavaScript runtime that support many 3rd party open source libraries like zlib, SQLite, NSPR, libiconv, libTomCrypt, OpenGL, ...