Highlights from Build 2017

I tend to describe Microsoft’s Build conference as a bit like Christmas for developers who use Microsoft’s tools and technologies to build software. This year was no exception – and there was plenty to be excited about.

As per usual, there is a vast amount of content published on Channel 9, most of which I have not gone through yet, but here are some of the top announcements that interested me the most:

Microsoft’s democratised AI offerings continue to grow and improve customisation

Microsoft have been promoting their Cognitive Services for a while now, and they’ve been getting more and more robust over time, now with 29 services up and running and available for developers to use.

One of the most exciting additions this year is the trainable image services. Being able to train AI to spot certain attributes on images is something that can have a huge impact on some of the technologies I build professionally.

The addition of Cognitive Services Labs allows developers to try out more experimental AI services, including Project Prague, a gesture recognition service.

It’s also worth mentioning that Satya said that, as solution architects and software developers, we should take accountability for the algorithms and experience we produce. We should be building inclusive systems which help empower people – in a way that they can trust. I agree with him.

Azure Cosmos DB is a shiny new multi-model global scale data service

As well as bringing much needed MySQL and PostgreSQL service offerings to the cloud, Microsoft have also announced their latest home grown cloud-native database service, Cosmos DB.

As a software architect, having Cosmos DB will allow me to make much better choices about the consistency of data solutions I am designing without having to worry about indexes or where the data will rest at run time.

The global distribution of Cosmos makes it a lot easier to make ensure that the data is as geographically close as possible to the end user. It’s essentially an extension of Document DB, but allows for a multi-model interface: key-value, column family, graph, and document.

As Cosmos DB is built on the Document DB technologies, there is already an emulator which can be used locally at development time. For me, this is a must when choosing cloud technologies.

For me the timing of the Cosmost DB announcement is really great, as a planet scale database is something I’ve actively been looking at for a new project I’m working on. I’m looking forward to learning more about it.

New tools for Azure developers and administrators

Azure is becoming one of the most important assets that Microsoft has. It’s the centre of many of their initiatives including AI, IoT, microservices, and more. Their continued work to strengthen this platform has made it easier than ever for developers to get up and running with all of these new services through a coherent set of tools and development kits.

New tools like the Cloud Shell and the Azure Mobile App are part of this. Unfortunately for me, the PowerShell version of the Cloud Shell isn’t available yet, nor is the Windows version of the Mobile App. However, the improvements to the Azure CLI are most welcome. Under the covers the Cloud Shell uses the shiny new cross platform command line interface for Azure and is already logged in and configured, making it super easy to get up and running. I’m a huge fan.

We’re still missing an Azure desktop app though – I still think there’s value in having a version of the Azure portal that doesn’t require using a web browser. Using Electron is probably the best way for Microsoft to achieve this and I’m unsure why they’ve not already provided a desktop app.

A powerful new feature called Snapshot Debugger will integrate with Visual Studio to make debugging production easier than it ever has been. You can create snap-points on certain lines of code which will instruct Azure to collect information as the application is used. It’s very impressive, and doesn’t affect people using the production application in any way.

I’m keen to try this out but it seems like it is going to be a powerful new way to fix issues in production without the security risks involved in pulling production data to a developer’s local machine for debugging. Awareness of production data is a must for companies who use customer data, and tools like this will help with adherence to the Data Protection Act and security standards like the popular ISO 27001.

Microsoft has a new mantra

A clear message from Build 2017 was that developers shouldn’t be placing all of their business logic and intelligence inside Microsoft’s cloud infrastructure, instead they should be considering how devices on the edge of this cloud could be leveraged to improve the solution.

Intelligent Cloud and Intelligent Edge

Not only does this make more sense, but it’s also something that Microsoft is uniquely positioned to provide. As a long term supplier of back office / on premise software, they’ve already got a foot in the door of many companies data centres. Improvements to Azure Stack and Azure’s IoT offerings allow logic to be moved between Azure’s cloud, to on premise data centres, and even to embedded edge devices.

Azure IoT Edge. is an example of how logic can move between the cloud and edge devices through a single management infrastructure:

  • Run AI at the edge to reduce latency and allow for offline-scenarios
  • Perform analytics and proactive decisions at the edge
  • Move logic from cloud to edge at any time
  • Management of edge devices from a central location
  • Simplify development
  • Reduce bandwidth costs

While these tools are very interesting to me, I have a feeling we’re still a little way off. The innovations here are huge and not to be taken lightly and I expect more to come over the next few years.

Cortana and Bot Framework improvements

One of the more obvious changes is that Cortana has come out of the phone itself and she’s now coming to other devices like the Harman Kardon’s Invoke intelligent speaker. (Yes, this counts as an intelligent edge device!)

General improvements have been made around the Bot Framework too. It’s now easier than ever to use natural language for common actions like taking payments from users.

Cortana Skills have been created to better link Cortana with services built on the Bot Framework and Adaptive Cards make it easy to write interactive cards which work across all platforms.

These integration improvements aside, I’m not convinced Cortana herself is moving fast enough and I’ll have to write up some more of my thoughts in a follow up to last year’s thoughts.

Windows 10 Fall Creators Update

Aside from the stupid name, it looks like there has been a steady progression for the Windows 10 platform.

The update brings a number of much-anticipated features including a cross-device clipboard, pick up where you left off, OneDrive on demand sync, and much more.

One of the best new features was the timeline view, which shows previously used applications across multiple machines. I’m not sure how well this will work for me, so I’m looking forward to getting my hands on it so that I can try it out.

Interestingly, the addition of a few apps to the Windows Store have caused quite a commotion:

  • iTunes – a must-have for iPhone users will be coming to the Windows Store. I don’t use it, but I understand the gravity of what this means to users and the pressure it will apply to Google to bring their apps to the store too.

  • Linux – we’ve had Ubuntu for a year, but now the Windows Subsystem for Linux has been updated to include Fedora and SUSE. Who’d have thought it would be Microsoft to really bring Linux to the desktop?

One of the more impressive apps was Windows Story Remix (the video is worth watching!), which takes advantage of many of the platform and service offerings to include an impressive experience for users who want to create video content with their photos and videos. While this isn’t something I do very often I certainly appreciate how well Windows Story Remix has been executed.

The fall update also brings the long-awaited replacement for the Metro design language…

Fluent Design System 😍

Microsoft’s design system has had a rocky past due the company being forced to drop its “Metro” identity early on in life, and it has hobbled along with the less memorable “Microsoft Design Language” since before Windows 10’s introduction.

Finally, they’ve sorted themselves out and come up with a new name for their design language.

While it is an evolution of the existing Metro principals (see my previous rundown), the new design’s new direction takes into account five key areas:

  • Light
  • Depth
  • Motion
  • Material
  • Scale

Fluent Design is something that really interests me, so I’m going to write more about this in an upcoming post.

Developer Tools, New APIs and much more…

It’s no surprise that there have been a load of improvements around the developer tools and other services too:

  • Visual Studio 2017 for Mac
  • 3rd party integrations for Microsoft Teams
  • .NET Standard 2.0 and XAML Standard 1.0
  • Azure Functions Improvements
  • Much more…

Exciting times!

Highlights from Build 2016


Even though I have never attended a Microsoft Build conference in person I always learn so much from them.

Every year there are new platforms to try, lots of documentation to read, and many presentations and recoded sessions to watch.

I still have a lot of videos to watch, but here are some of top announcements from Build 2016 which matter to me the most as a developer.

Windows 10 as the best OS for Developers

A number of new features coming to Windows 10 in the “Anniversary” update were shown in the day-one keynote, and then even more features where shown at sessions throughout the conference. Solid improvements to the inking, biometrics, and the Action Center were all well received.

Windows Ink

Many of the features shown help fix minor annoyances in the system. For example, pressing on a live tile showing a preview of a news article can now take you directly to it, and notifications dismissed on the PC or tablet will automatically get dismissed on the phone too.

One of the most exciting new features was the addition of Bash (on Ubuntu) on Windows which is both technically very interesting and extremely useful for many development workflows. The new Ubuntu subsystem will allow any (command-line) Linux application to run natively on Windows. This instantly unlocks a massive amount of tools and utilities for developers, making common scenarios significantly easier from Windows.

Bash on Windows

As a huge fan of command line interfaces I’m going to go into this in more detail in a future article – but essentially Microsoft are positioning Windows to be the ultimate developer platform, no matter what operating systems you use for your solutions.

Azure is growing up with more data centres and services

Microsoft would prefer you use Azure when you deploy your applications though, and the day-two keynote showed that is still serious about the cloud.

Improvements which interested me the most included Azure Functions, Service Fabric, Containers, DocumentDB, and much, much more.


Azure is the future of Microsoft, and by the numbers they going strong. They’re expanding their datacentres and really betting big on the cloud. This is no surprise to Microsoft watchers, but it’s good to see steady improvements here. Many of which I will use.

Visual Studio keeps getting better

I spend absolutely huge amounts of time in Visual Studio so any improvements here have a very positive effect on my productivity.

Visual Studio 2015 Update 2 was released (with lots of improvements) and an early preview version of Visual Studio vNext was also shown. I’ve tried both and they’re definitely going in the right direction for me.

Visual Studio

I’m especially looking forward to some of the improvements coming in the Visual Studio installation experience moving forward. This should make setting up new development environments much faster, and the side-by-side installations means there’s much less risk when installing previews.

App development for Windows, iOS, Android

The mobile app development story from Microsoft is stronger than it ever has been. This year brings a number of improvements to the Universal Windows Platform (UWP) itself, and a more integrated store experience which now includes the apps on the Xbox One and HoloLens.

The Desktop App Converter lets you wrap up existing Win32 and .NET apps into UWP packages, allowing access to new features like UWP APIs – including Live Tiles. Even though I don’t currently develop any Win32 or .NET applications that I want to put in the store, this is an important step and I’m looking forward to the benefits of this as an app user.


For targeting non-Windows devices, the Xamarin platform is now the obvious choice. After recently purchasing Xamarin (and their amazing talent) they’ve decided to make Xamarin available for no extra charge with Visual Studio. And that includes shipping it with the free Community version. Very cool.

The combination of UWP and Xamarin means I can directly apply my C# and .NET skills to making applications for a wide range of platforms, sharing many code components. It’s really coming together nicely.

.NET and the continued move into Open Source

As well as making Xamarin’s development tools free to Visual Studio users, the folks over at Microsoft also announced their intention to open source the Xamarin SDK (including the runtime, the libraries, and command line tools), and give the governance of it over to the .NET Foundation.

Mono, the cross platform and open source sibling of the full .NET Framework has also been re-licenced to be even more permissive, and given to the .NET Foundation. (To be honest I actually thought this was already the case!)

.NET Core, the future replacement of both the .NET Framework and Mono, also saw steady improvements – my favourite of which was official F# language support:

$ cd hellofs/
$ ls
$ dotnet new --lang f#
Created new F# project in /home/julian/hellofs.
$ # I can now dotnet restore and run this F# app using .NET Core!

The Future of Cortana and Conversation as a Platform

So far everything I have mentioned has been mostly around solid updates to existing platforms, but this year’s Build included a slightly different way of thinking about productivity with the idea of Conversation as a Platform.

Conversation as a Platform

The Microsoft Bot Framework provides templates for creating bots with C# and JavaScript, as well as connectors to simplify their interaction with services like Slack and Skype. When linked with the new Cognitive Services, these bots can understand natural language and perform tasks for the user.


The demonstration of talking to Cortana through Skype was very interesting – where essentially Cortana can act as a broker between the user and other bots on the Internet which can act as experts in their field. I found this very compelling, and something I can see myself using.

As this is as subject that interests me greatly, I’ll be writing more about this over the next week or so.

And everything else…

Of course, there’s no way I could summarise everything I looked at so I have skipped a number of cool announcements ranging from Microsoft Graph to HoloLens.

The hard-working folk over at Channel 9 have videos for many of the events and topics, so be sure to check them out if you’re interested. I’m very thankful that these videos are all made available for everyone to watch, I really enjoy watching them.

Interactive showdown: csi.exe verses fsi.exe

Visual Studio 2015 Update 1 brings with it a nice little utility called C# Interactive which you can access from the View menu, but we’re going to have a look at the command line version which you can run from the Developer Command Prompt for VS2015.

Using this tool you can quickly run C# using a REPL environment – which is a fast and convenient way to explorer APIs, frameworks, and script existing code from .NET assemblies in a non-compiled way.

While this is new to Visual Studio and C#, we have enjoyed this functionality for a long time with F#.

(Technically this has been around for a while, but it’s officially shipping now!)

I decided to take a look at the new csi.exe application, and compare it to how I already use fsi.exe and see if it’s enough to make me switch my default command line tool.

C# Interactive

For me the most important way I’d use C# Interactive is via the command line, so it’s important to know what it’s capable of, even though you may not need to use the advanced features right away.

To find out the current version and get a list of the command line options in C# Interactive, just add the /? switch and read the output:

PS> csi /?
Microsoft (R) Visual C# Interactive Compiler version
Copyright (C) Microsoft Corporation. All rights reserved.

Usage: csi [option] ... [script-file.csx] [script-argument] ...


/help Display this usage message (alternative form: /?)
/i Drop to REPL after executing the specified script.
/r:<file> Reference metadata from the specified assembly file (alternative form: /reference)
/r:<file list> Reference metadata from the specified assembly files (alternative form: /reference)
/lib:<path list> List of directories where to look for libraries specified by #r directive. (alternative forms: /libPath /libPaths)
/u:<namespace> Define global namespace using (alternative forms: /using, /usings, /import, /imports)
@<file> Read response file for more options
-- Indicates that the remaining arguments should not be treated as options.

Form a first look, I can see that csi.exe has all of the command line options I really want in normal use – I especially find /i to be useful – but we’ll come to that shortly.

F# Interactive

F# Interactive has been around for a lot longer, and is built on different technology under the hood – so there are a more options going on here, but we can take a look by providing a similar -? switch:

PS> fsi -?
Microsoft (R) F# Interactive version 14.0.23413.0
Copyright (c) Microsoft Corporation. All Rights Reserved.

Usage: fsi.exe <options> [script.fsx [<arguments>]]...

Input Files

--use:<file> Use the given file on startup as initial input
--load:<file> #load the given file on startup
--reference:<file> Reference an assembly (Short form: -r)
-- ... Treat remaining arguments as command line arguments, accessed using fsi.CommandLineArgs

Code Generation

--debug[+|-] Emit debug information (Short form: -g)
--debug:{full|pdbonly} Specify debugging type: full, pdbonly. ('full' is the default and enables attaching a debugger to a running program).
--optimize[+|-] Enable optimizations (Short form: -O)
--tailcalls[+|-] Enable or disable tailcalls
--crossoptimize[+|-] Enable or disable cross-module optimizations

Errors and Warnings

--warnaserror[+|-] Report all warnings as errors
--warnaserror[+|-]:<warn;...> Report specific warnings as errors
--warn:<n> Set a warning level (0-5)
--nowarn:<warn;...> Disable specific warning messages
--warnon:<warn;...> Enable specific warnings that may be off by default
--consolecolors[+|-] Output warning and error messages in color


--checked[+|-]Generate overflow checks

--define:<string> Define conditional compilation symbols (Short form: -d)
--mlcompatibility Ignore ML compatibility warnings


--nologo Suppress compiler copyright message
--help Display this usage message (Short form: -?)


--codepage:<n> Specify the codepage used to read source files
--utf8output Output messages in UTF-8 encoding
--fullpaths Output messages with fully qualified paths
--lib:<dir;...> Specify a directory for the include path which is used to resolve source files and assemblies (Short form: -I)
--noframework Do not reference the default CLI assemblies by default
--exec Exit fsi after loading the files or running the .fsx script given on the command line
--gui[+|-] Execute interactions on a Windows Forms event loop (on by default)
--quiet Suppress fsi writing to stdout
--readline[+|-] Support TAB completion in console (on by default)
--quotations-debug[+|-] Emit debug information in quotations
--shadowcopyreferences[+|-] Prevents references from being locked by the F# Interactive process

As you can see there’s a lot more options for F#, but many of them are not needed for every day use.

Quick Interactive Use

It’s fairly common that I use F# Interactive just to test out how part of the Framework behaves.

In this instance, I’ll use HttpUtility.HtmlEncode method to see see what output I get when one of my emoticons is encoded into HTML-friendly characters.

PS> fsi

Microsoft (R) F# Interactive version 14.0.23413.0
Copyright (c) Microsoft Corporation. All Rights Reserved.

For help type #help;;

> open System.Web;;
> let encode s = HttpUtility.HtmlEncode(s);;

val encode : s:string -> string

> encode "<(>_<)>";;
val it : string = "&lt;(&gt;_&lt;)&gt;"

This is how I’d do it in F# – we could call the HtmlEncode function directly, but creating functions is so easy with F# that we might as well shorten the name to make it nice and easy if we need to run it multiple times.

The function encode actually returns a string rather than printing it to the screen, but F# is setting that output to a special value called it – a special identifier which is used for displaying the value of the last expression on the screen. It’s handy, and you’ll see why.

Alright so here’s my first attempt to do something similar in C# Interactive.

PS> csi
Microsoft (R) Visual C# Interactive Compiler version
Copyright (C) Microsoft Corporation. All rights reserved.

Type "#help" for more information.
> using System.Web;
> HttpUtility.HtmlEncode("<(>_<)>");
(1,1): error CS0103: The name 'HttpUtility' does not exist in the current context

Ah. HttpUtility is missing because it hasn’t loaded the clases from the System.Web.dll assembly. I didn’t notice on the first line becuase of the way namespaces work – the namespace exists, but not the class we want. No problem, we just reference it using #r – you reference assemblies this way in F# too!

> #r "System.Web"
> HttpUtility.HtmlEncode("<(>_<)>");

This worked and we have access to the static HttpUtility class and the HtmlEncode method – however the output has not been displayed to the screen because C# Interactive doesn’t have that the specal it value F# had.

I didn’t realise this at first but in the absense of the it value F# has, the C# Interactive prompt introduces a slightly different syntax for when you want to see the value.

> HttpUtility.HtmlEncode("<(>_<)>");
> HttpUtility.HtmlEncode("<(>_<)>")

Notice the difference a semicolon makes? This is important, and something I missed when first trying out C# Interactive. Avoiding the semicolon would normally result in invalid C#, but this is a great way to view the output as if you’re typing it into the Immediate Window in Visual Studio.

Let’s also create a function using normal C# syntax so that we don’t have so much typing to do. Notice that I’m going to call this function without the semicolon so that I can see the output.

> string encode(string s) { return HttpUtility.HtmlEncode(s); }
> encode("<(>_<)>")

Loading Scripts

Let’s keep things simple, we’ll take the functions we just created in each langauge, and create a script file so that they can be loaded up when we start an interactive session.

First of all, let’s do it with F#. Here’s the content of encode.fsx:

open System.Web
let encode s =

And then we can run it from the command line using the --use switch. This will drop us into an interactive prompt after the code file has been loaded.

PS> fsi --use:.\encode.fsx

Microsoft (R) F# Interactive version 14.0.23413.0
Copyright (c) Microsoft Corporation. All Rights Reserved.

For help type #help;;

val encode : s:string -> string

> encode "<(>_<)>";;
val it : string = "&lt;(&gt;_&lt;)&gt;"
> encode "<(^o^)>";;
val it : string = "&lt;(^o^)&gt;"
> encode "<(T_T)>";;
val it : string = "&lt;(T_T)&gt;"

Not bad at all. So let’s do the same thing with the C# interactive, using a file called encode.csx:

#r "System.Web"
using System.Web;
string encode(string s)
    return HttpUtility.HtmlEncode(s);

I love that they used a similar extension! And again, we can run the code file and then get an interactive prompt as above using the /i switch.

PS> csi /i .\encode.csx
> encode("<(>_<)>");
> encode("<(^o^)>");
> encode("<(T_T)>");

We have the same end result, though like before the actual functions behave slightly differently. C# Interactive gives a cleaner output here, though you can always clean up the F# Interactive prompt a little bit by using the --nologo switch.

Use Inside PowerShell

Because I want to get access to both of these utilities as fast as possible, I have added a few lines to my PowerShell profile which will ease their use.

I’ve mentioned doing this kind of thing before – and I highly that developers using Windows spend a good amount of time learning PowerShell – but here’s a little snippet that may be useful.

$PROGFILES32 = "C:\Program Files (x86)\"
# create an alias to the full path of the executable
Set-Alias fsi "$PROGFILES32\Microsoft SDKs\F#\4.0\Framework\v4.0\fsi.exe"
Set-Alias csi "$PROGFILES32\MSBuild\14.0\Bin\amd64\csi.exe"
# add helpers which include common switches
function fsu ($fsx) { fsi --nologo --use:$fsx }
function csu ($csx) { csi /i $csx }

Adding this to my profile means I can just run them using fsu encode.fsx or csu encode.csx respectively. Very easy.