Hosting an infinite number of apps in the cloud for free, on your own domains.

Longer title: Building and hosting an infinite number of scalable secure web apps on custom domains, with no vendor lock, only using skills I already have, for free.

Sometimes I talk myself into the seemingly impossible. While writing a stubborn paragraph of my book (Your First Product... go sign up!), I wanted to know just how cheaply I could host a custom web app in the infamous cloud.

For example, could I do it for free?

Cheap is good, cheap is wonderful, but FREE is magical. If you can host apps for free, then you can host as many apps as you dream up. Whoosh! Bing! Blam! Another thought? Another app!

Once I'd realized what I was attempting to achieve, I put it in tweet form as this:

And promptly gave up on the idea, as it was clearly BONKERS. Or at least I tried to give up on it. My mind kept turning the idea over... there must be a way to have it hosted for zero dollars, without vendor lock in.

Sometimes if you set up a system with a lot of constraints, you see that it's impossible. And you then need to work out which constraint to relax. (This is called "engineering")

So I did what any true engineer does, I used a spreadsheet. I put all the constraints in their own column, and listed different solutions on each row, then checked which constraints they broke.

Anything involving Azure cost too much, and was too likely to lead to vendor lock in. Heroku had a lot going for it, but relied on technology I don't use. GitHubPages had a lot of advantages -- but don't permit any kind of server-side code. At the same moment as I started looking into 'Parse' I heard that it had been shutdown.

Finally I stumbled on an architecture that suited my needs. It was a hybrid, like that mythical beast with the head of a lion and the belly of a zebra, or however those old myths used to run. They were pretty popular back in the day.

Here's what I came up with.


That's right. As befits a zero-dollar architecture diagram, I have resorted to the use of Comic Sans.

The front end is static html and javascript, hosted by GitHubPages. GitHubPages are free, and they let you configure your own domains or sub-domains, so for example I could have "" be served from GitHubPages, provided I own the domain (...and have wrestled it back from some pesky Ukrainian). One downside is that the repository has to be public, as private repositories at github still cost a little money, but since this is just the front end of the website, I see no harm in making the code available.

The front end uses javascript (json) to talk to the back-end, which is an app hosted elsewhere. The back end is hosted by appharbor using their free tier. Because it's the free tier they won't let me have a custom domain, they assign a url such as but that's okay: this is just a back-end which the customer never need see.

AppHarbor is a great way to host the back-end of a site, because I can deploy to it from the commandline, by just pushing to a repository. I don't want the backend code to be public, so I need a private repository. Private repositories at github cost money. For one low monthly price you get unlimited private repositories... but that's not good enough for my constraints, nuh uh. Fortunately, bitbucket gives you unlimited private repositories for a single user, and appharbor integrates just as nicely with bitbucket as they do with github.

So there we have it. GitHubPages front end, appharbor+bitbucket back end. A way to host an unlimited number of small applications, on custom domains, in the cloud, without learning any new tech or getting slugged with fees from anyone.

(One thing I didn't solve was the 'secure' part. I wanted to have the whole thing encrypted end to end, using the mystical glory of Let's Encrypt, a new certificate authority who are ushering in a golden age of https everywhere. The most commonly suggested way to do this is to use cloudflare with a custom domain in front of github pages, though there's also a kloudsec solution. I haven't tried either, so can add nothing of value on the topic. The 'free tier' of AppHarbor includes 'piggyback SSL' which I think might be sufficient for this example.)

And here's the app:

Guess A Guid ← Can you guess a GUID? Try it and see for yourself!


How to Left Pad, for real

So someone removed a bunch of their packages from the node package manager, and this in turn broke a lot of other people's software builds.

There have been 1 million articles written (so far) wherein sweaty-fingered coders tie themselves in predictable knots asking:

  • Does this mean NPM is doomed?
  • Does this mean opensource is doomed?
  • Does this mean opensource wins, because it can respond so quickly?
  • Does this mean micro-dependencies are terrible?
  • Can something still be a knee-jerk reaction even if someone specifically says it's not?
  • Does this show that NPM is an evil corp?
  • Doesn't this mean that you should do a trademark check before publishing anything ever?
  • Should you check in your dependencies? Should your dependencies have checked in their dependencies?
  • Has everyone forgotten how to code all of a sudden?
  • It's like everyone has gone crazy! Has everyone gone crazy!?

...and so on. But I don't want to ponder any of that.

I just want to look, very carefully, at the code itself, in the center of this maelstrom....

function leftpad (str, len, ch) {
  str = String(str);
  var i = -1;
  if (!ch && ch !== 0) ch = ' ';
  len = len - str.length;
  while (++i < len) {
    str = ch + str;
  return str;

I've had left-padding (and right-padding) functions on my brain lately, as they were added to the most recent release of NimbleText. I was curious if this function behaved the same as mine. 11 lines... what could possibly go wrong?

So I grabbed this implementation and tested its behavior.

I was surprised to see it gave different results to my function!

Specifically -- what does the leftpad function do if you give it a string such as 'HELLO' and ask it to left-pad it to a width of 4 characters (i.e. a length that is smaller than the initial string)?

In NimbleText, to answer this question I asked my customers, who uniformly pointed to the behavior of Oracle's LPAD function.

Oracle's LPAD function, if given a len that is smaller than str, will truncate the result.


LPAD("HELLO", 4) returns "HELL".

So that's what I implemented for NimbleText.

But that's not what this function does!


leftpad("HELLO", 4) returns "HELLO".

The difference is minor -- but minor things can have dramatic consequences.

For example if someone assumes that leftpad(someString, 10) has an invariant property that it always produce a string that is 10 characters long, they could soon end up with a security vulnerability.

I mentioned this on Twitter and celebrity whitehat hacker 'OJ' responded with:

I wouldn't want a leftpad() function to trim strings

Which I attribute to a latent desire he has to see more and newer vulnerabilities in code (not that there's any foreseeable shortage of vulnerabilities looming otherwise)

...but anyway -- what would you expect from a leftpad("HELLO",4) ?

Should the package manager maintain a running vote, and the people can decide democratically on every question?

Or should there be... I dunno... what's the dirtiest word in software... A standard?


Today I Learned

A month ago I saw an article on Hacker News about someone who created a github repo, in which they recorded any interesting and resuable solutions to the problems they encountered.

I started doing the same thing, storing little markdown files inside my 'utils' repo (the private repository where I keep copies of all the little tools I use on every machine, plus my powershell profile etc.)

Pretty quickly I had 100 such files, and I looked into what I could do with this growing knowledge base. I found out about 'gitbook' which is a way of rapidly turning a github repo full of little markdown files into a genuine book.

So here's the product... the free book I wrote without trying to write a book...

today i learned title image

Today I Learned (

(and here's the git repo, error corrections welcome!)

Inside that book, I've written a short article on "getting started with gitbook" so you can do the same thing, including details on how to use gitbook locally (for example for documentation inside the enterprise!)


npm is not just for node... npm is for EVERYTHING

npm -- the node package manager -- is a bit of a beast. In fact it's a lot of a beast. it's a crazed, snaggle-toothed snarling monster hell-bent on ruling the entire world and crushing all opposition beneath its giant cloven hoofs.

If you've only heard of npm in connection with node, you may be excused for thinking, as the name implies, that it is just for managing node packages. node is a specific server-side technology, so perhaps that is all that npm cares about?

Not at all. npm cares about everything. npm does everything.

Here is how I thought of it, when I was young and foolish.

Consider these three basic types of package managers:

  1. Machine-level package managers
    These package managers install entire applications. Think chocolatey, apt-get or yum.
  2. Application-level package managers
    These package managers install components an application relies upon during dev/build. Think nuget (or gem, or npm)
  3. Front-end package managers
    These specialize in components for the front end component. Think Bower. (Or nuget in previous versions)

I thought npm comfortably belonged in that second tier. It's like nuget I thought, but for node. Shrug.

Then I learned about the "-g" (global) parameter. And realized that with npm you can install packages that are available "globally".

So it has crept up into that top tier as well. It can do what chocolatey, apt-get and yum do.

Not a problem I thought... then I took a closer look at Bower. It's an interesting and well-liked package manager for the front end.

But I started hearing more and more about people preferring to use npm for front end packages too. Apparently you combine npm with Browserify and/or dedupe -- and you quickly see it's got the 3rd category covered as well.

Not to worry I thought. npm is just one part of a rich tapestry of tools.

Task runners, for example, are another category of tool used by web developers. There's a growing number of these task runners out there, and two in particular always seem to be vying for supremacy: Grunt and Gulp. But recently I've been hearing more and more about npm scripts.

"npm scripts" are a way of making npm aware of easy to use aliases that point to longer scripts, exposed by a package.json file.

They give you a convenient way to just use npm for everything.

And that's where we've come to now. You can just do every damn thing in npm. I'm kind of hearing an update to Atwood's law, something like:

Everything that can be done in npm will be done in npm. Any thing that can't be done in npm will be replaced by a thing that can.

Maybe npm stands for "node's perfect monster".

Even if you have never used node, and never want to use node, you will still need to use npm.

By January of 2017, all other technologies are being phased out permanently, in order to make room for nothing but pure npm.

Introducing intellisense-like command completion for npm on windows

I say all of this by way of introducing a helpful tool that gives you intellisense-like behavior when using npm on windows.

Doug Finke has written a powershell module that gives you tab-completion when using npm on windows.

In powershell version 5 from an elevated prompt, find and install the NPMTabCompletion module, from the new Powershell gallery, like this:

Find-Module TabExpansionPlusPlus -repository PsGallery | Install-Module -Scope AllUsers
Find-Module NPMTabCompletion -repository PsGallery | Install-Module -Scope AllUsers

(As shown, you first install TabExpansionPlusPlus)

Then, from a freshly opened prompt (that does not need to be elevated), import the module into your session.

Import-Module TabExpansionPlusPlus
Import-Module NPMTabCompletion

(You will also need to either add those commands to your profile, or run them in every powershell session where you want npm tab completion to work.)

This not only gives you intellisense-style completion for npm commands, it also looks in the nearest package.json file to see if there are any commands (i.e. npm scripts) you've specified locally.

Here's an animated gif of it in action:

npm tab completion

I've published a more thorough guide to installing it, here: npm tab completion with powershell. There are troubleshooting notes as well, in case you fall into the same problems I fell into (including the sad news that the time came to uninstall PsGet which has been replaced with PowerShellGet)

(That documentation is at a "Today I Learned" site I've been building out, day by day. There's a lot of stuff there already. I'll blog about how to build a site like it as soon as I get a moment!)


Console Is Forever

When I was a wee-little lad, in the 80's, I loved the console.

What a mystical conversation: human talking directly to machine!

When "Windows 95" was released I thought the console was dead.

When I started "working" as a professional developer, I sometimes used the console, but considered it a quirk due to my peculiar upbringing.

As the WindowsTM versions rolled by, I never shrugged off this "quirk". One day "PowerShell" (Monad) arrived — I adopted it, unlike a lot of my colleagues. I used it on and off.

One day, a colleague said "You use the console a lot... are you a linux developer?" Interesting I thought.

Recently I've looked at linux and npm.

I realise now that my "quirk" — my penchant for consoles — is not in "spite" of GUI superiority.

It's a much simpler reason.

2-D GUIs are nice. One day they may be 3D! (That would be fun.)

But consoles... The good old console. Console is forever.


The Secret Life of Connection Strings In Oracle! (Oracle DBA's Hate Me.)

I don't know if you've ever had to work with Oracle, but if you have, and if you're primarily a .net developer like me, I wonder if your experience is anything like my own.

The very first thing you do with Oracle, as a developer, is to connect to it. And you soon discover that the way connections are specified in Oracle is batshit insane.

With a SQL Server connection string, there are a couple of quirks*, sure. But essentially, the connection string holds all of the details that are used to connect. Every detail you could possibly want goes into the connection string.

(* Regarding connection strings. I've always wanted to register the site "" which simply redirects to "" but is much much easier to remember.)

In Oracle, a connection string is just the first mirror in a long hallway full of mirrors. A hallway that leads into a maze of mirrors, that takes you to a forest filled with monsters and more mirrors.

We had an issue recently where a connection string that worked perfectly well on 7 different machines did not work on an 8th machine. When we looked into how the connection worked on those first 7 machines, we found it was different on every single one. When we took this knowledge to the 8th machine, we began to feel a terrible sense of foreboding. The further into the issue we looked, the more terrified we became. Eventually the authorities found our skeletal remains at the bottom of a steep ravine clinging to a well thumbed printout of Oracle documentation.

Here is how you usually connect to Oracle:

You follow some simple instructions, and you get some generic error messages. You're not panicking, you're just a little out of your depth. Eventually some oracle DBA scoffs into the room (that's the way they walk, they scoff as they walk) pushes you off your keyboard and does some magic incantations in front of your computer. You keep looking over their shoulder, trying to learn about this strange magic. You see strange registry keys, shocking environment variables, keywords like "tnsnames.ora" and some other disturbing flashes of technical wizard-pokery. Was that LISP!? For a moment you get one sickening glimpse into the infinite pit of sorrow.

After that it just works, and you soon forget how tricky it was, you confine it to a dark corner of your psyche. The rest of your experience with Oracle is never as daunting as that first connection. But by then the Oracle DBA has cemented their role as an irreplaceable technical God.

Well I'm here to blow the lid open on their whole racket.

I'm here to teach you all the things they don't want you to know. I'm going to expose their little shitshow and all of its pathetic quirks.

Get ready.

In an Oracle connection string, the Data Source attribute is used, not for connecting to remote machines, but for searching your local machine. I'll give you an example. If the connection string looked like this:

User Id=scott;Password=tiger;Data Source=bbSales12

What do you think happens with that Data Source value? (bbSales12)

Here's what happens....

The value of Data Source is an Alias. And Oracle wants to Resolve that alias.

According to the documentation, Oracle first looks in "the connection pool" to see if it has already "Resolved" that alias.

Given that, at this point, we haven't started talking to any remote machine, we must be referring to some kind of local connection pool. Probably some kind of in-memory dictionary. Documentation is sketchy of what that is. But let's press on. Ignore that bit.

Assuming we haven't resolved this alias before, then we need to look at tnsnames.ora.

What is tnsnames.ora I hear you ask?

It is the third mirror. It is magic. It is a file that contains enough configuration detail that your machine should be able to connect to the remote machine.

It is likely to be this file that the administrator was futzing with. (Or some other things that will be explained shortly)

First up: how does Oracle find the tnsnames.ora file?

To find this magical file, Oracle relies on an environment variable, TNS_ADMIN. That is the second mirror.

If you don't have this environment variable set, then Oracle will not be able to find the tnsnames.ora file (or the sqlnet.ora file, which will be described later)

With powershell you can look at all your "environment variables" like this:

Get-ChildItem Env:

Get-ChildItem has these two aliases that you might find more comfortable: dir, ls,

So you can type

dir env:


ls env:

And to see value of the TNS_ADMIN environment variable, type $ENV:TNS_ADMIN

Now, assuming you have that variable set, it will show you a path. Follow that path, and look for a file named tnsnames.ora

(Some people will tell you that tnsnames.ora is located in ORACLE HOME\NETWORK\ADMIN. This is confusing advice that you can ignore. More on that later)

Was tnsnames.ora there? Good. Now look inside. It is a text file. So you open it with sublime or vs code, or whatever the cool kids are using at this time of day.

You should find the alias defined via a piece of configuration language like this:

  (ADDRESS= (PROTOCOL=tcp)(HOST=bb12Sales-server3)(PORT=1521))

What is that language? It looks a bit like lisp, or a bit like JSON. It sounds a lot like Greenspun:

Any sufficiently complicated program contains an ad hoc, informally-specified, bug-ridden, slow implementation of half of Common Lisp.
Greenspun's Tenth Rule

I haven't been able to find any official name for this syntax, other than "the syntax of the .ORA files". It's used by Listener.ora, tnsnames.ora, sqlnet.ora, protocol.ora and tnsnav.ora

(By the way, whitespace is insignificant, but recommended.)

Now -- what if you don't have a tnsnames.ora file? You could of course go ahead and create one, then try and put the right protocol, host, port and servicename into it.

But there's another intriguing possibility.

Way back at the start of the process, inside your connection string, you can embed one of these weird lisp-like strings directly.

Instead of

User Id=scott;Password=tiger;Data Source=bbSales12

You can have:

User Id=scott;Password=tiger;Data Source=
        (ADDRESS= (PROTOCOL=tcp)(HOST=bb12Sales-server3)(PORT=1521))

How's that for a snappy connection string that rolls off the tongue. In official parlance what we've done is add a "connect descriptor" to our connection string.

Now you must be wondering.... what other syntaxes can this Oracle connection string withstand?

The answer is "pretty much all of them."

As an attempt at irony the wise folk of Oracle invented something called "the Easy Connect Naming Method".

This lets you specify a Data Source like this:


For example:

"user id=scott;password=tiger;data source=//bb12Sales-server3:1521/"

Now that looks much simpler doesn't it? Why am I stating that this is ironic? Because of this little footnote from the documentation:

Prior to using the easy connect naming method, make sure that EZCONNECT is specified by the NAMES.DIRECTORY_PATH parameter in the sqlnet.ora file.

This introduces the "sqlnet.ora" file which is also searched for in the TNS_ADMIN folder.

Inside that file, if you specify


...then you'll be able to use "the Easy Connect Naming Method".

If you're brave you can also add LDAP and ONAMES into the mix. But that's outside the scope of this rant.

This EZCONNECT syntax really is a bit easier than the alternatives, so I don't think their attempt at ironic naming was a complete success ;).

Oracle Home.

Now I mentioned that some people will tell you that tnsnames.ora is located in ORACLE HOME\NETWORK\ADMIN

I'll cover what that means, but first I have to tell you, No! tnsnames.ora is not necessarily located in ORACLE HOME\NETWORK\ADMIN. It can be located anywhere! That's what the TNS_ADMIN environment variable is for: for allowing it to be located anywhere, irrespective of where "ORACLE HOME" is. (Your friendly system admins may locate tnsnames.ora on a network drive where it is easier to administer en-masse.)

Now what is this "Oracle Home" business, hmm?

That is the third mirror down the first hallway on the right. It is a dead end. But I will tell you about it anyway.

When (or rather "if, against all common sense") you installed Oracle, it will have been installed into a location that it refers to as the "Oracle Home".

It's possible that there is an environment variable called "ORACLE_HOME". But failing that -- the value of "Oracle Home" may be stored in the registry.

Where in the registry, exactly? Any one of these places my poor dear friend.


But as long as TNS_ADMIN is set correctly, you don't need to worry about those registry entries, or the absence of an ORACLE_HOME environment variable.

Okay, that's all I wanted to say about Oracle connection strings. In order to cleanse your mind of this horrible mess, I suggest you learn about how .net locates an assembly at runtime or look into error handling with MVC.


The little known beauty of .ensure files

Ever broken the build?

Everyone who uses a revision control system has, at some point, accidentally committed a file that should never have been committed. For example, a ".dll" or a dreaded ".suo" file.

To save us from this entire category of problem, the ".ignore" concept was created, and implemented as a built-in feature of many revision control systems (for example ".gitignore" and ".hgignore")

Similarly: everyone has, at some time, forgotten to "commit" a file that other parts of the code relied upon. Thus, the build has been broken, time and time again. Throughout the world, builds are breaking at this very moment, for the lack of some file that could've so easily been committed.

There is a widespread but little known feature, as simple as the ".ignore" file, that addresses this problem.

If you're not familiar with it, let me tell you about the ".ensure" file.

A ".ensure" file informs the repository of a few basic standards of file hygiene that must be achieved before a push can be issued.

Much like a ".ignore" file, You can copy a ".ensure" from some one else's similar project into your own, and everything will just work.

You can copy a ".ensure" from some one else's similar project into your own, and everything will just work.

For example, if someone else has written a C# based project, where the ".ensure" file says, in effect:

"If there is a '.sln' file in the repo, each referenced project file must also be part of the repo 
"If there is a '.csproj' file in the repo, each referenced file or resource must also be part of the repo

...then you can copy their ".ensure" file into your own C# project, and you will be protected against so many entirely predictable build breakages.

Admittedly, the syntax of a ".ensure" file is a little baroque, cramming up to four "selection" DSL into the one handy file format (regex, xpath, jsonquery and globs).

But most people never write them: they simply copy them from place to place. (Copying files comes naturally to developers at every rung of the ladder.)

And testing the effectiveness of a given ".ensure" file is very easy (you just try to break the build with a missing file). So the community easily converged on a working set of files for any given project type.

The only problem that anyone has with ".ensure" files, is that they are something Leon simply dreamed up, five minutes ago, while doing the dishes. Implementations are yet to exist.

I for one say: go for it.


The Laggard's Guide to Getting Started with 5 Using Yeoman.

Note: 5 has been renamed to ASP.NET Core 1.0. The instructions below are unaffected. 5 represents a complete re-invention of the .net platform. The foundations have shifted completely. Throw out everything you ever thought you knew about computers, and get ready to have your hair blown back, your face melted and your whole outlook on life re-defined.

If you are anything like me then you've let much of this stuff zoom past you over the last year. Well it's time to get up to speed my friend.

Here are the steps to get up and running with 5, using Yeoman. In summary:

  1. Optional: Install vs2015 including "Microsoft Web Developer Tools"
  2. Install Asp.Net 5
  3. Use dnvm to upgrade your dot net versions and set active version
  4. Install chocolatey
  5. Install node.js, using choco (in order to get npm)
  6. Install visual studio 'code'
  7. Install yo, using npm
  8. Install generator-aspnet, using npm
  9. Generate a new site, using 'yo aspnet'
  10. Visit the new site in vscode
  11. Restore all nuget packages for the project
  12. Serve the website locally using dnx web
  13. Browse to http://localhost:5000

And now I'll go through them in slightly more detail -- but still very quickly.

If you want to go through this in a more comprehensive manner, I recommend this pluralsight course "Building a Web App with ASP.NET 5, MVC 6, EF7 and AngularJS by Shawn Wildermuth".

1. Optional: Install vs2015 and make sure to include "Microsoft Web Developer Tools"

We're not going to be using vs2015, but using vs code instead. Still I expect the installation gods will devour your soul if you don't install vs 2015 before installing 5. So consider doing this first.

  1. Install Visual Studio 2015
  2. And be sure to specify that you want the "Microsoft Web Developer Tools"

Come back when you have that done, a few hours later. If you're living in Australia where our "internet" is made from a few frayed old copper strands, add one extra month.

2. Install Asp.Net 5

This step, and all subsequent steps, are mandatory.

  1. Go to Get.Asp.Net
  2. Click on "Install For Windows".

If you're not on windows, then follow for the other instructions.

Alternatively, read the detailed install instructions.

3. Use dnvm to upgrade your dot net versions, get new versions and set active version

Once is installed, you can run dnvm. dnvm is a commandline tool for managing which version of the .NET runtime to use.

At a command prompt, type each of these commands in turn:

dnvm upgrade
dnvm install 1.0.0-rc1-final -arch x64
dnvm install 1.0.0-rc1-final -r coreclr 
dnvm install 1.0.0-rc1-final -r coreclr -arch x86
dnvm use 1.0.0-rc1-final -r clr -arch x64 -p
dnvm alias default 1.0.0-rc1-final -r clr -a x64
dnvm list
dnx --version

These commands, in order, will:

dnvm upgrade:
Install the latest 32 bit version of the runtime, make it active and assign the default alias to point at it

dnvm install 1.0.0-rc1-final -arch x64:
Install the 64 bit version of the common language runtime (clr)

dnvm install 1.0.0-rc1-final -r coreclr:
Install the 32 bit version of the "core" clr

dnvm install 1.0.0-rc1-final -r coreclr -arch x86:
Install the 64 bit version of the "core" clr

dnvm use 1.0.0-rc1-final -r clr -arch x64 -p:
make the 64 bit clr "active", and persist this setting. (i.e. not just for the life of this console session)

dnvm alias default 1.0.0-rc1-final -r clr -a x64:
apply the alias "default" to the 64 bit clr.

dnvm list:
list all of the dot net versions.

dnx --version:
show the version of the dot net execution environment.

You don't need to install all of those clr's. But knowing that you can is fundamental to the shift that is occurring.

(I don't know why .net is now back at version 1.0. Perhaps they forgot to drink their ration of the semantic versioning kool-aid.)

4. Install chocolatey

You may already have chocolatey, in which case, skip this step. Otherwise. From a console window with administrative privileges type:

@powershell -NoProfile -ExecutionPolicy Bypass -Command "iex ((new-object net.webclient).DownloadString(''))" && SET PATH=%PATH%;%ALLUSERSPROFILE%\chocolatey\bin

Chocolatey is now installed on your machine and available in your path.

If you're using an OS other than windows you probably have a package manager already, with which to perform the next step.

By the way: Rob Reynolds who maintains Chocolatey: what a hard worker! Thanks for your efforts Rob!

5. Install node.js, using choco (in order to get npm)

Use chocolatey to install node.js. We're doing this in order to get npm, so that we can get yo, so that we can generate aspnet projects at the commandline.

Yes -- it is a convoluted process. But don't let that stop you. Just do it once and get on with your life.

Here's a simplified diagram that might help:

Sorry, wrong diagram, this is the one I meant:


choco install nodejs

You now have node.js and with it, npm.

At this point you can add "rockstar web developer" to your json resume.

To test out node.js, type "node" in a console. Now enter:

console.log("hello from node");

And you should see "hello from node" echoed to the screen. Exit node with ".exit". Or by hitting Ctrl-C a few times.

6. Install Visual Studio 'Code'

Visual Studio Code is a cross-platform code editor optimized for building and debugging last year's crop of web and cloud applications. In 6 months time it will probably be abandoned on the side of the information superhighway, but for now it works beautifully.

  1. Install VSCodeSetup.exe

Or follow the windows install instructions.

7. Install yo (using npm)


yo is the scaffolding tool that is part of Yeoman

You install it using npm (the nodejs package manager) that was installed as part of node.js, above.

npm install yo -g     

The "-g" here means: 'global'. So it is available from future console sessions. Not unlike the "-p" above.

8. Install generator-aspnet (using npm)

yo's purpose is to "scaffold" applications, i.e. construct the bare-bones of an application, that serve as a starting point for further development. Think 'Dos On Dope', but with slightly larger concept count.

Yo doesn't actually do the generating itself -- it gets "generators" to do all the actual work. This seems like a pretty good deal for Yeoman who outsources all of his work but still takes the credit. For example, to scaffold an aspnet5 application yo needs the aspnet generator. Download it, using npm, like this:

npm install -g yo generator-aspnet

Now you have all of the pieces in place and can build a website. Try not to burst with the anticipation that pushes all of your blood into your head.

(To see a list of yeoman generators available from npm, type: npm search yeoman-generator, and to see the generators installed locally type: yo --help)

9. Generate a new site (using 'yo aspnet')

Simply type:

yo aspnet

Press "3" to select "Web Application".

Type the name of the name application, e.g. "HelloFromYo".

Press enter. A whole lot of very interesting text will scroll by in a flash.

At the end of it, you'll have a directory structure like this:

    ¦   +---Account
    ¦   +---Manage
    ¦   +---Account
    ¦   +---Home
    ¦   +---Manage
    ¦   +---Shared


10. Visit the new site in vscode

from the console type:

cd HelloFromYo
code .

And visual studio code will open the current folder.

Otherwise, from within vscode choose "open folder" and navigate to the newly created website's root folder.

11. Restore all nuget packages for the project

Within Visual Studio Code, navigate to any of the code files within the project.

VS Code will soon wake up and notice that the relevant nuget packages are not available.

A note at the top of the screen will say:

"There are unresolved dependencies from 'project.json'. Please execute the restore command to continue."

Click "Restore."

This will call the dnu commandline utility, to download the nuget packages used by the project. (You could've performed this yourself using 'dnu restore' and 'dnu build')

It will download something like 265 nuget packages. Because "componentization" is the special something that has been missing from your life all this time.

Your generated web app is now ready for launch. Be excited.

12. Serve the website locally, using dnx web

From the console, while in the root folder of the webapplication, type:

dnx web

This will launch the kestrel webserver, hosting the website. You will see from the console the exact port number, but it will probably be 5000.

Why does it launch the kestrel webserver? Because in the file "project.json" under "Commands" it says the Web command uses "Microsoft.AspNet.Server.Kestrel". I mention that, just to give you a taste of the magic.


13. Browse to http://localhost:5000

Browse to http://localhost:5000 to visit the site.

And there it is, in all its stupendous glory.

Now sit down and have a long think about the good old days, when all you needed to do to get into a development environment was restart the Commodore 64 and type '10 print "hello"[enter]20 Goto 10[enter]Run'.

Then stop thinking about all that and start looking through the code to see how it works. There's some fascinating magic in there to be sure.

Better learn it quick, before it's time to throw it all out again and rewrite everything in $NextLang.


Persistence Framework for Accepting Nonsense.

This site uses cookies. By browsing this site you accept that the site uses cookies.
—One million different sites.


"Unauthorised reproduction of this DVD is bad. Opinions expressed do not represent the views of the yada-yada, and you can't show this DVD on a prison oil-rig etc."
—Every time I put in a new Breaking Bad DVD.


There should be just one check box.

Do you get it?
Don't ever ask me again. EVER!

Perhaps at the moment someone turns 18, a representative of the bureaucrat-class could show up at your house with a clipboard and ask:

"Do you get it?"

You sign something, just once, and the lawyers can leave you alone with their repetitive questions.

Perhaps if you paint the top of your door frame with lamb's blood, they'll just tick the "Gets it" box and move on without interrupting your day.

I can't fix the DVD problem, but the cookies on the web problem is easy to fix.

At a browser level, we, the computer peeps, could agree on a micro-format. It's about time we did something useful with the semantic web!

Those nagging cookie questions could have an attribute, a data-cookie-nag='true'. And if, in about:settings you've set the flag for 'I accept that the internet uses cookies and i know how to inspect and or disable them', then the browser will quietly delete that node from the DOM. (Or if it's a form with a single checkbox, check it and submit it).

I see no possible problems or roadblocks to this idea being immediately implemented world wide.

It starts with you. If you are being forced to implement one of those cookie-nag checks, then you can add the micro-data today. There's nothing stopping you.

Once you've done it, you can inform me here:

Once it's implemented on 1000 sites, I'll let you know, so you can patch chrome and firefox to take advantage of it.


Buying a car: A text adventure game.

You are standing on the street. Before you there are three car dealerships, with a sign on top of each one. One says:
"MAZDA", one says "HYUNDAI", one says "SUBARU".


You walk into the Hyundai dealership.

> Look

You look around. There is a car and a car salesman.

> Look car

You move toward the car, hoping to look at it. The car salesman detects your presence and moves between you and the car.

"Hi there, I see you're looking for a car there. Got any questions, just ask me!" 

The salesman blocks your path.

> Ask Salesman.

The salesman doesn't know what you are asking.

> Ask salesman about car.

The salesman doesn't know anything about "car"

> Ask salesman about price.

The salesman doesn't know anything about "price"

> Ask salesman about car.

The salesman has no fricking clue as he is a complete idiot.

> Leave.

You leave the car dealership.

You are standing on the street. Before you there are three car dealerships, with a sign on top of each one. One says:
"MAZDA", one says "HYUNDAI", one says "SUBARU".

> Quit

You cannot quit. You must acquire transport.

I hate this game.