Random Thoughts

AngularJS by Example - My new book on AngularJS

2015-04-01T09:49:00.000+05:30

For some time now i have been blogging on my corporate blog instead of my personal one. But this is something that i should share on my personal blog too as it is big :)

I have a book out there "AngularJS by Example"! Published by Packt this book targets beginners to intermediate audience.

Head over to my blog post here to learn more about the book.

You can also checkout and buy the book on Packt website.

Happy reading!

AngularJS validation directives library

2014-03-10T09:31:00.003+05:30

I wrote a set of directives that make validation in AngularJS a tad simpler.

The directives are available at

https://github.com/technovert/angular-validation

I have detailed about the directives on my company website

http://blog.technovert.com/2014/03/angularjs-form-validation-library-directives/

and there is also a jsfiddle to play with

http://jsfiddle.net/3vxgy/1/

or

Common Pitfalls using AngularJS

2014-02-12T11:40:00.001+05:30

I recently blogged about some common pitfalls using AngularJS on my companies blog. Anyone interested in the details should visit my post at.

http://blog.technovert.com/2014/02/common-pitfalls-angularjs/

You can also subscribe to my feed here

Happy Reading :)

Integrating ASP.Net MVC with AngularJS

2013-12-05T09:59:00.001+05:30

We recently released a Project Template for ASP.Net MVC and AngularJS here and here.

I did a follow up post detailing how the project template has been setup and how to make AngularJS and MVC play well together on my company blog.

If you are interested in understanding how the template works head over to my blog post and read it!

Visual Studio Project Template for AngularJS SPA + ASP.Net MVC \ Webapi

2013-09-23T18:07:00.000+05:30

I am very exited to announce that our team at Technovert has release a Visual Studio project template for AngularJS. More details on how to install and use this template are available on the Technovert Blog

Knockout.js has been a standard library that .Net devs have been using for creating rich interactive business apps. But AngularJS from Google provides a more complete framework and is gaining a lot of dev community support. This project template can help you get up and running with AngularJS in no time and evaluate it for your future needs.

AngularJS Model - ViewModel

2013-08-06T17:27:00.002+05:30

Recently Robin Ward did a comparison between AngularJS and Ember (both the post and the comments are interesting read). One of the complaint that Robin had that AngularJS does not provide any specialized model class. As described in the blog

There is no standard Model base class, nor is there a component or interface in AngularJS that defines what a model is supposed to be.

What that means is that anything defined on $scope can acts as model, as this sample on the blog describes

For some (like me) this is the flexibility where as for others who want a more structural approach this is a nuisance.AngularJS does not even provide any definite guidelines around how the model should be actually designed to be effective with AngularJS

This leads to what we call as $scope pollution. We define multitude of properties on the $scope object and managing it becomes difficult. Looking at the code one cannot tell what is the actual model and what properties are just to support UI logic.

Model Design - Convention

We have been using AngularJS for some time now and we have devised a convention which has helped us to keep $scope pollution under control. Here is how we organize our model.

On the Controller scope we create a object viewModel. ViewModel according to us is a model which has been tailored to support a specific UI, helping in easing data binding.
We define a property model on either the controller scope or on the viewModel defined above. This model is something that is loaded from the server and updated back on edits.

This is how a sample controller looks

So any controller defines only one or two property directly on the $scope and that's it !

What could be the advantage of this small convention? Many :)

Readability

To start with, anyone reading the code can know what the model is and what variable are just used to support AngularJS data binding (such variables are created on viewModel object).

Ease of passing model around

Any decent size angular app/page would consist of multiple partial views and directives working together to provide the necessary functionality. One of the challenge in such setup is how to keep partial views and directives as independent as possible.

For partial views (views loaded using ng-include) here is how we define our template and main view.

As you can see, any template that requires a model to work on, gets it through ng-init. Just looking at the ng-init assignment one can determine what does a partial template require. Passing model around to these sub views become easier. Each of these partials can have their own controller and create their own viewModel object.

For directive, this is what we can do

For directives it becomes easier for us to create isolated scope, which has great benefits in terms of re-usability.

As you can see in both cases dependencies are clearly defined and there is a single root object at all levels.

IIS Url Rewrite and HTTP POST data

2013-05-03T12:32:00.000+05:30

If you play around with IIS Url Rewriting rules and try to do redirects on an HTTP POST you loose your POST data.

To make sure you get the post data you have to set the `redirectType` to `Temporary` within your rules. So the action configuration looks like this

<action redirectType="Temporary" type="Redirect" url="http://{HTTP_HOST}{REQUEST_URI}"> </action>

You may think what scenario warrant a POST request redirect. We faced one such scenario while doing SSO with a federated Identity Provider (IP) such as Google, Azure AD. In federated authentication scenario once the user is authenticated by the IP, it redirects back to your site with claim tokens in a POST request over secure channel (https). In our case we wanted to redirect to user back http after receiving the request. But any redirects were causing loss of token. By doing a 307 (Temporary) redirect we were able to preserve the post data.

Remote login on Azure Roles

2013-05-02T10:04:00.000+05:30

In case you do not follow this guy :), there has been one significant change to Azure which would make developers life much easier while debugging and testing an azure deployment.

You can now enable remote login into azure deployment without requiring you to enable it during the package creation. This means even if the deployment happened with Remote login not enable, you can still enable remote login from azure management portal directly. More details are available here

So no more forgetting remote login option. Just do a deployment and if required enable remote login from azure management portal.

Logging .Net Exception Details with Complete Stack

2013-01-18T17:34:00.001+05:30

I learned today that in .Net doing a ex.ToString() would log the complete exception details including any inner exception details. The important thing here is that it would do this recursively so if there is an InnerException it would call ToString() on it, which in turn can call ToString() on the next level hence forming a recursive chain.

So we can do away with such format string

System.Diagnostics.Trace.TraceError("The error was {0}. Stack Trace {1}. InnerException Details {2}",ex.Message, ex.StackTrace,ex.InnerException.Message);

Google Driven Development GDD

2013-01-10T19:45:00.001+05:30

In this era of TDD, BDD, DDD and all other *DDs, we have a breed of developers doing Google Driven Development (GDD). Sounds interesting, indeed it is :).

Here is how to identify a GDD guy

This person does not believe in writing anything on his own, copy/paste from blogs, articles, source code is a way of life. And since I have not written it I don't care how it works as long as I see something working. His best friends are Ctrl-C, Ctrl-V and Google.
Cannot write 5 lines of code or code for 5 minutes without doing a Google search.
His productivity is directly related to performance of corporate network, and how fast Google searches. If the network is down, it's better to go for a coffee.
Search hours on Google for "Object reference not set to an instance of an object" and then believes that the whole world is facing the same issue. Who cares what the context of problem is.
To solve the error

He then tries each and every solution that Google GOD provides.
If by fluke he is able to fix or camouflage the issue, he considers himself something special, but in his heart he knows he has no clue what happened. Thanks Google for saving his day and continues his pursuit to become a better GDDer.
If not, then it's time for the Lead, the next GOD in Line. The Lead should be able to fix this issue, What else are Leads there for :)

Reading documentation is so very lame ! Documentation is for sissies, real developers do Google.
Religiously believes that all results returned by Google are correct. Google has employed a huge workforce just to make sure results are correct.

Feel free to comment and tell me more such traits of a GDDer and i would add to my post :)

I may have exaggerated a bit, but what to do, I am seeing such people all too often.

Azure and Package Deployment for Specific Roles

2013-01-06T16:10:00.000+05:30

When a single Azure Subscription Service has multiple roles (Web and Worker) in it, updates to the role cannot be selective. Time and again I have had scenarios where I wanted to update the one role but did not want to touch the other. But as Azure deployment works, there is a single package and it updates all roles in the subscription.

While doing an update today I was pleasantly surprise that now while updating the site I do have an option to select which specific role to update. Here is what the dialog looks like now

So now we can update specific roles without touching other ones. Great!!!

Things to keep in mind here would be, once deployment is complete you cannot tell which role was updated and which is still running on a older version. But the deployment name does change. So i think this is a good feature for some quick fixes, else we should stick to simultaneous updates only.

Angularjs and inplace edit

2012-12-15T16:06:00.000+05:30

AngularJS is an awesome framework and Angular Directives make this framework super awesome :)

AngularJS directives help us extend the existing HTML DSL. You can create new tags, extend functionality using existing tag and create reusable component.

I recently created a Inplace editable control in AngularJS. It basically allows to do in place editing for elements like H1, H2, H3 div etc, basically all read only control. I took the initial idea from this fiddle and added

Text selection on edit.
Cursor positioning
Handle the Enter key press.

Here is my fiddle

Integrate ASP.Net MVC + WebAPI with Autofac in 5 mins

2012-12-02T13:40:00.000+05:30

Recently i configured Autofac with ASP.Net MVC 4 and WebAPI. Here are the steps required to configure Autofac

Install package Autofac.MVC4
Install package Autofac.WebApi
Add a class AutofacBootstrap. Use this class to configure the dependencies within your project.
Open Global.asax file and add a method like
Call the method from Application_Start
And you are done !!

Appharbor build notification using Google Talk (XMPP) - Implementation

2012-09-16T16:12:00.001+05:30

In my last post about creating an Appharbor notification bot, I described about the problem domain, the solution approaches and how the end to end scenario works.This post would dig deeper into the implementation aspects of the solution.

To start with we need to understand the two esential ingredients for this recipe

Appharbor Service hooks: Appharbor Service hooks are the building block for the solution. Appharbor provides capability, where it can post to any public url with status of a build on completion. Service hooks are available for events

Build Success.
Build failure.

XMPP protocol: XMPP is a protocol used for Instant Messaging (IM) communication. It allows us to relay messages to the users IM client using XMPP servers (in our case Google Talk messaging servers).

With the understanding on these two concept we are ready to dive into the nitty gritty of the solution.

Assuming that the user did the initial configurations to setup notifications (see my earlier post), here is what happens behind the scenes when a build notification is triggered from Appharbor

One may find it odd that parts of the solution are hosted on AppHarbor and parts on Windows Azure worker role.

The reason for this was, inability for IIS to host the XMPP client and allow it to keep persistent connections to XMPP server. Whenever i was trying to instantiate a XMPP client (agsXMPP) the call silently failed when hosted in IIS. In a console app it worked fine. I did not dwell much into the issue but based on my initial research the XMPP client tries to create a persistent connection to XMPP server (in our case Google Talk) but IIS does not allow this.

Since AppHarbor provide more of a web endpoint (presumably in IIS) the solution did not work when hosted in AppHarbor(even in Azure webrole). I had to create a Azure Worker Role and expose a web endpoint from the worker role. Using ASP.Net Web API Self hosted feature, I was able to host a http endpoint within the worker role, but it did require some configuration changes for the worker role.

To know more about the implementation details look at the source code available on GitHub.

Project Appharborbot.Webhost contains the bot web site code.
Project Apphraborbot.Worker contains the Azure worker role code.

Overall this was an interesting exercise which showed how multiple technology\tools can be combined to create a quick n smart solution.

Appharbor build notification using Google Talk (XMPP)

2012-07-22T14:47:00.000+05:30

Appharbor recently ran a contest to encourage developers to use their API and showcase the API capabilities.

Based on some ideas floated by the Appharbor team I too decided to build something useful\interesting for Appharbor platform.I decided to create a desktop notification client which can notify user when the build is complete, its status in terms of success or failure.

How Appharbor works is that you push your code to a git repo supported by Appharbor (github, bitbucket etc). As soon as the checkin is done, Appharbor pulls the code base build it, and updates the running website. It's like a single click deployment!

I started to explore my option for a desktop client and some of the options I though of were

Build a desktop windows app and integrate with the Appharbor API. This approach was the most flexible approach. I have full control of features and capabilities, but would take a good amount of time to develop.
Use third party desktop notification apps like Growl for Windows and use the Growl API to integrate with Appharbor. The shortcoming with this approach is that the user needs to install Growl for Windows on his machine, but it is better than the first approach as i can make use its existing notification capabilities

What i finally decided was to create a solution that uses Google Talk as the notification client. The advantages of this approach were

Easy to build and within a short time span.
No extra installation required. Now a days most people have these messenger clients installed.
Cross platform \ Cross browser compatible. Works on Mac, Linux, Windows i.e on any system that supports a XMPP client.
It seems cool :)
This approach can support any XMPP based client.Google Talk being one of them. This wiki entry details about the existing clients supporting XMPP.

The disadvantage with this approach is that you are limited by Google Talk + XMPP protocol capabilities.

So after dabbling for few days i created a working prototype. Here is how it works.

User creates an Appharbor Application.
Navigates to http://appharborbot.apphb.com/ to generate Service hook url.
Generates the Service hook url from this site.
Adds the generated Service Hook url to his Appharbor app under the Service Hook section.
Adds contact appharbornotificationservice@gmail.com to his Google Talk
Receives all build notifications in his Google Talk!!!

The final outcome, user now gets all build completion and failure notifications. This is how notifications look

I would be soon publishing the second part in this series, where I would talk about the technical aspect of the solution, which is as interesting as the solution. This solution is an amalgamation of Windows Azure and Appharbor so just watch out for my future post.

Azure Blob file download not resuming !

2012-06-19T18:30:00.000+05:30

Azure Blobs are a scalable infrastructure to host\store files. I was working on understanding the behavior of large file uploads and downloads and create a reasonable strategy to support both scenarios in our application.

Azure blob file can be downloaded using any browser or internet download manager. For a public blob the url of file would suffice, but for downloading a private blob one needs to create a url containing the Secured Access Signature.

While testing downloads I realized that in case I pause and resume the download in Firefox or Chrome or any download manager the download restarts !!! I was expecting that the download should start from the place it paused. Imagine the scenario where we have uploaded half of a 2GB file and resuming the download would start over again !!! Clearly the Azure Blob storage was not supporting HTTP Range header.

As it turns out the Azure blob service is not a trivial file server. It is more of a Storage Application Server. It exposes an REST API to interact with it. The ability to support Range header(hence resume) was added in a later version of Blob API.

The default API version that we interact with is 2009-09-19. This is the version used in case we do not specify any (version can be added in HTTP header x-ms-version). Since the browser does not send any custom version header while making the GET request we do not get the resume support.

The fix here was to update the blob service properties for our account using the Set Blob Service Properties API method. This method takes a parameter DefaultServiceVersion. This parameter defines the default version of blob API to use in case API version is not specified in the HTTP header. Setting this value to 2011-08-18 fixed the resume problem.

Now the browser was able resume broken downloads and download managers could even download in multiple chunks to improve download speed.

I did not find any option in the Azure Management portal that allows us to set the the above property so I had to do it using the REST API. I am sharing the code that I used to update the API version below

Alogorithm: Finding Longest Increasing Subsequence using Patience Sort in C#

2012-05-24T18:32:00.000+05:30

Finding the Longest Increasing Sequence is an interesting problem and there are multiple solutions available with varying time complexity. The solution that i plan to share today is the uses of Patience Sort to find such a sequence. The solution finds a particular longest increasing sub-sequence as the original sequence may contain more than one such sequence.

In 1999 The Bulletin of the American Mathematical Society published a paper by David Aldous and Persi Diaconis entitled: “Longest Increasing Subsequences: From Patience Sorting to the Baik-Deift-Johansson Theorem”.

I have implemented this solution in C# and derived it from this excellent article which provides a Python implementation for the same.

This is how patience sort works

Take a deck of cards labeled 1, 2, 3, … , n. The deck is shuffled, cards are turned up one at a time and dealt into piles on the table, according to the rule
A low card may be placed on a higher card (e.g. 2 may be placed on 7), or may be put into a new pile to the right of the existing piles.
At each stage we see the top card on each pile. If the turned up card is higher than the cards showing, then it must be put into a new pile to the right of the others.

I have left out the part that returns the sorted list of numbers as it is not relevant for the longest sub-sequence problem.

The first thing to note here is

Once the cards\numbers are organized into piles. The total number of piles denote the size of longest sub-sequence.

To get the longest sequence some extra book keeping is required. According to the paper

Lemma 1. With deck π, patience sorting played with the greedy strategy ends with exactly L(π) piles. Furthermore, the game played with any legal strategy ends with at least L(π) piles. So the greedy strategy is optimal and cannot be improved by any look-ahead strategy.

Proof. If cards a1 < a2 < … < al appear in increasing order, then under any legal strategy each ai must be placed in some pile to the right of the pile containing ai-1, because the card number on top of that pile can only decrease. Thus the final number of piles is at least l, and hence at least L(π). Conversely, using the greedy strategy, when a card c is placed in a pile other than the first pile, put a pointer from that card to the currently top card c′ < c in the pile to the left. At the end of the game, let al be the card on top of the rightmost pile l. The sequence a1 ← a2 ← … ← al-1 ← al obtained by following the pointers is an increasing subsequence whose length is the number of piles.

What this means is whenever a card is added to a stack (except first stack), the card should keep a back pointer to the topmost card on the pile on the left.

To get a better understanding of the above logic, i have created a diagram detailing the links that get created during the patience sort process for the list of numbers

31,16,7,39,5,12,32,18,9,1

The longest increasing sub-sequence using the above algorithm is

5,12,18

The gist for the implementation is available here

Caching Images downloaded from web on Windows Phone Isolated storage

2012-05-13T18:55:00.000+05:30

I was helping on a Windows Phone application where the requirement was to cache the images the phone downloads on the isolated storage for offline viewing.
I wanted a solution which was simple and as transparent as possible. While researching I found someone wrote a Silverlight converter for loading images from isolated storage. Taking that as a base I created a converted which can

Load image from web (http + https), and persist it to isolated storage.
In case of network connectivity issues can load the same image from isolated storage. It does that by mapping the http url to a isolated storage location.
In case the network is down and the image is neither there in cache, loads a default image, passed as parameter to converter.

Here is the gist for the implementation.

To use the converter

Import the name space.
Declare the converter as resource.
Set the Image Source Property to use this converter like this

Architect? Give me a break :)

2012-04-24T23:24:00.000+05:30

Davy Brion has posted some rants about the Microsoft MVP program and the type of people that are attracted to it.
What David has mentioned holds true in some other areas of IT industry too.
"Architect" another cool but severely abused title :) If you take David's post and replace "Microsoft MVP" with the word "Architect" you would find everything still holds true. I am a strong believer that "Architect" as a designation\title should be done away with.

In my 8+ years of IT experience I have seen quite a few of this breed :) With exception of few I found most of them just good with

UML modelling tools (such as Visio)
Generating high quality project artifacts (Such as Function Specs, Design documents)
Buzz word mongers.
Smooth talkers

These maybe important qualities to have in an Architect but a sound technical background is a non negotiable. Everyone knows what is the end result if you engage such people. Such Architects mostly produce things that look impressive on paper but implementing them most often then not becomes infeasible.

Well, its not the Architects fault always. In a typical engagement this is how the Architect works. At the start of the project the Architect

Understands the problem domain.
Talks to the client coming up with a solution approach.
Shares the solution approach with the client and key stakeholders.
Gets their feedback.
and finally creates artifacts for the team to get started.

What starts of as a full time gig for the Architect, gets reduced to a advisory role as the project progresses. And after some time the Architect is gone, entrusting people like us the responsibility of implementing their grand vision. This is where I believe the problem lies.

When we are working in an Agile age where requirements evolve, designs evolve, technology evolves, mandating a design at the start of the project is incorrect. The time where Architectural decisions are made, are times of maximum uncertainly as the project is getting off ground. What ends up happening is that Architectural document\guidelines gets sidelined and the solution evolves based on what the immediate requirements are. This is shear waste of resources.

There is another problem with this approach. Where is the accountability? If I am not going to develop something I am architecting \ designing, I can virtually propose anything. Literally anything !! If things work out fine I can always take the credit else I can always pass the blame to others. Works well for me! I can keep on doing that project after project and keep going up in the organization ladder, with virtual little or no real contribution to any project I work on.

At the end I am not going to present any solution for this situation but point my readers to an excellent presentation by Simon Brown where Simon discusses the role of the software architect, challenging some of the current assumptions and trying to redefine it in the context of Agile development.

So you got a job offer !!!

2012-04-05T20:23:00.002+05:30

So you got a job offer !!! It always a great feeling irrespective of ones experience level. You start envisioning about your fatter pay-cheque, your role, your growth prospects, may be a new city (in case you are transferring). During this euphoria we may throw caution to the wind and not do a thorough check about the organization, type of work, work culture. The consequence of our carelessness can cost us, sometimes dearly.

Through this post I just want to highlight things one should consider while looking out for jobs and while accepting any job offer. Since i am an IT professional i can and would only talk about IT companies and jobs.

Product Company vs Services Company: Many may not have any preferences on this but this distinction has an affect. Services company is all about billing and one almost always works under deadline pressure, delivery pressure. The story is different in product companies where activities revolve around the product. Time to market is important in product development organizations too but not as important as services organization.

Company's IT division vs Pure IT company: If you are driven by technology, IT division within a large organization may not suite you. One needs to understand, non IT companies are driven by business needs not technology needs. As always there are exceptions to this rule too, but one should be aware of this.

Background Check: Background check is very important before you accept any offer or join a company. Try to extract as much data about the company's background, the type of work it does, the work culture, company policies. Some of the avenues for such research are

Your Friends: Check with you friends, and try to see if someone has some experience with the company. Many a times you friend may get you connected to someone who can help you better in the research.
LinkedIn: LinkedIn is a professional network, where one can find about any company, it's current and previous employees. There is a wealth of information available on LinkedIn. You can look for your connections within the company. You could find 1st, 2nd or 3rd level connection, so go ahead and leverage these connections. Try to gain insight into the company. If possible find connection who have been past employees with the company.
Company Review Sites: I know about two such sites www.jobeehive.com and www.glassdoor.com. Check if someone has posted reviews about the company. Number may not be reliable but if there is a general trend of low numbers and bad reviews, that is a clear RED flag

Interview Process: The complete interview process is about selecting\rejecting a candidate. The company assess a prospective candidate through these interviews. What we may not realize is that, other way round is also true. Interviewing is a good time for us to judge the company as well. While interactive with the company employees we can get a fair idea about their talent base, their culture. Some of the things that should make you circumspect

Having things too easy: A good company tried to push the candidate to his limits. They don't want to make things easy for you. Always strive to work with people who are smarter than you. If the interview was too easy for you that may mean you are over qualified for the job.
Having Single Technical Round: Unless you are directly interview by Director, CEO, CTO etc, single technical round selection is strictly no no.Good companies do not rely on single person feedback. The more the merrier.

This is what I have currently. I would continue to update this list over time. I want this list to be exhaustive collection of tips\links that highlights what to check for in ones prospective employer.

Case of missing Iron Foundry VMC client

2012-03-26T12:17:00.000+05:30

While installing the latest Iron Foundry VMC client and Cloud Foundry Explorer there is an installation issue. Due to this VMC client install gets lost if both the VMC Client and CF Explorer are installed irrespective of the location of the install.

As it turns out, if you install VMC client first and CF Explorer next, installer would delete the VMC install directory.

If you install CF Explorer first and try to install VMC client later you get this error

"This version or newer version of VMC Command Line Tool is already installed."

Current workaround would be copy the installed folder for VMC before starting CF Explorer install.

Hope it helps.

NoSQL Data Modelling

2012-03-25T19:12:00.001+05:30

Recently in one of the project we planned to use some NoSQL Document database. One of the reasons we though document database would be a great fit was that we could get started without setting up any DB schema and start shoving entities into the document database.

Nothing can be further from the truth. Data modeling is as essential and as fundamental an exercise for NoSQL stores as it is for RDBMS. It is just that the concepts are different as compared to RDBMS where normalization\de-normalization provide some guidance. In my pursuit to understand how to model data for a document database i came across some great reference material that i want to share here.

MongoDB Schema Design : Great resource from MongoDB. Start with this one first.
MongoDB Data Modeling: A quick read but explains some important concepts related to modelling. Embedding vs Linking.
MongoDB Data Modeling and Rails: Covers an example to make things more clear.
NoSQL Data Modelling Techniques: A great post that explains about data modelling techniques for all NoSQL type stores.

I believe this is enough content for one to have a decent understanding on the topic and get going with the real modeling.

Trying to make sense of REST (over HTTP)

2012-03-13T18:27:00.001+05:30

People who come from WCF/ Web Services background and transition to RESTish type communication infrastructure make mistakes that follow a common pattern. I have seen it with people who are new to REST and with people who have spent some time working with RESTful services. (Throughout the post where every I mention REST I mean REST over HTTP) The common misconceptions are

Not understanding the difference between HTTP POST and GET, in fact having too little or no clue about HTTP Verbs. For most devs interaction with server involves calling a client proxy method, handling the request on server, sending a response back from server, handling the response on client. This abstraction means no one knows what happens at infrastructure level. When programming for HTTP one cannot ignore the abstraction, else we cannot take full advantage of the medium.

The manifestation of the above affect is, either every call becomes a POST or every call becomes a GET.

When every call becomes GET. All CRUD are done through url. I have seen people trying to do crazy stuff with urls just to pass data from client to server.
When every call becomes POST (easy to do as compared to GET) , we fail to leverage the advantages of GET.

Assuming making HTTP call via GET\POST makes a service RESTful. This is a very common misconception and I too am culprit of thinking the same. But as I read more about REST and looked at examples I realize how far was I from a real RESTful implementation. I highly recommend REST In Practice book which is a great resource if one is serious about understanding REST.

Thinking the performance gains comes due to absence of SOAP envelop. The real performance gains come when data gets cached. This caching can occur at different locations starting from the servers network (reverse proxies), intermediaries, client network (proxies) and client browser.

Creating RESTful service and marking most of GET's as non cacheable. The common cause of this is not understanding the cache expiration in HTTP and therefore not taking advantage of it. Moreover during debugging cycles cached data is a pain to work with when one wants to test his\her fix.

These are some of the patterns that are at top of my head. What else can be added?

Contrasting Azure and IronFoundry Deployment Behaviour

2012-03-06T11:59:00.000+05:30

I have been exploring Iron Foundry a PaaS player in .Net space for the last few weeks. One of the things I liked about Iron Foundry was, how fast deployments happened. One of the reasons for this is how binary\file package gets uploaded to the cloud infrastructure. The Iron Foundry \ Cloud Foundry approach (since Iron Foundry is port of CF) of only uploading change-set instead of complete binary\file package makes this step of deployment quite fast.

What happens in Azure is

You create a package file (cspkg) using either Visual Studio (VS) or command line tool cspack
The generated package file contains the complete set of binaries and other resources such as images, css, javasripts etc
This package is then uploaded to Azure infrastructure and then deployed by the Fabric Controller.

In contrast the process that Cloud Foundry follows to decide what it should send to cloud follows a multi-stage process. When an application is ready to be pushed to Cloud Foundry the client and Cloud Foundry servers perform a number of steps to make sure only the required amount of binaries\files are uploaded. These steps include

Creating a metadata file.This metadata file contains Application Name, Url, Framework, Runtime, Instance Required, Memory reservation (this is similar to Azure configuration files cscfg)

Look at components (Decide what gems, libraries, npms, etc. are used in preparation for the next step.)
File based Fingerprinting (This is where a Manifest of files is built along with a SHA-1 hash associated with that file).
Send manifest to Cloud Foundry
The Cloud Foundry then responds with a manifest. This manifest lists only the files that Cloud Foundry needs, not all of the files that make up the application.
The client based on the manifest response creates a packages (compressed) which contains only the files that have changes.
The client then uploads the package

If you compare the process above the optimization that Cloud Foundry does as compared to Azure reduces the deployment time considerably.

With Azure I have time and again faced issues where we miss some file\code fixed and that required us to upload the complete package again. Imagine doing it for package of 200MB+ size on a slow network.

Windows Azure has improved this process a bit by supporting Web Deploy. This allows us to do incremental\differential upgrades as supported by Web Deploy model. But this approach again has it's own set of limitations as detailed here.

This posts only talks about the packaging and upload process for Azure and Cloud Foundry. If you want to know more about how Cloud Foundry deployment works look at these posts

All in all Azure can still improve this part of their deployment process to reap good performance benefits and improve development experience.

Nerddinner on Appharbor

2012-03-03T11:23:00.001+05:30

My last exercise on deploying Nerddinner on IronFoundry helped me a lot to understand the platforms capabilities. Time to do the same thing on AppHarbor, another PaaS player in .Net space. AppHarbor tries to emulate Heroku in .Net space. They are the 'Azure done right ' brigade.

As a developer I just needs to push code either to AppHarbor or some of the supported repositories such as CodePlex, GitHub, BitBucket and AppHarbor does the rest. AppHarbor

Builds the code.
Runs unit tests
Deploys the application on AppHarbor’s application servers (backed by Amazon EC2).

Other than that it also provides all the standard benefits as provided by most PaaS players.

To gain first hand experience on AppHarbor I decided to port Nerddinner onto AppHarbor. I already had the Nerddinner codebase in github so i could start right away.

Creating Account

The first step was to register my account with AppHarbor. Once registration is complete we get a single instance under the free tier. That was good enough for me to complete this exercise.

Linking to Repository

To use the AppHarbor infrastructure I had to link my github project with my AppHarbor account.

The first step here is to create an application from the AppHarbor dashboard.

Then link that application with the code repository (github in my case). I would not detail this process here as details are already available on AppHarbor. Things to note here are

AppHarbor repository are Git based repositories. Therefore if you plan to use AppHarbor repository you need to setup you git account locally, More details are available here
If you are using a third party repository such as CodePlex, GitHub or BitBucket you can look into the respecitve guides to help you integrate with you AppHarbor account.The guides are available at

BitBucket: http://support.appharbor.com/kb/api/integrating-with-bitbucket
Codeplex: http://support.appharbor.com/kb/api/integrating-with-codeplex
GitHub: http://blog.appharbor.com/2011/10/13/announcing-github-support

In my case since I was using my public github repository, hence went ahead and configured the link between my github repo and AppHarbor.

Deploying the Application

This is where AppHarbor shines.

I made some changes to my app on the local clone and pushed the changes to my remote github repository.

I went back to my dashboard on AppHarbor and I could see the application being built.

It failed for me for the first time due to some missing file but I was able to fix is promptly.

AppHarbor makes the build logs available for every build performed. These logs can come in handy while debugging the build failures.

I did a remote push again and the build kicked it again, this time succeeding. I fired the browser and navigated to the application http://nerddinner-4.apphb.com to verify the deployment. Nerddinner was up and running.

Not it was time for database integration.

Database Deployment
Nerddinner requires SQL Server to store its data and therefore I had to provision a SQL Server instance in AppHarbor.

AppHarbor provides free shared SQL Server instance add-on with 20 MB storage. That was enough for me to get started. I provisioned my SQL Server from AppHarbor addon page

Once provisioned the next step was to get the details of the new SQL Server instance. These details can be easily retrieved from the application dashboard. More details are available later in the post.

Since AppHarbor SQL Server instance is exposed to internet, I was able to connect to it from SQL Management Studio and execute the scripts for Nerddinner and aspnet DB (part of code-base).

Database was ready, application was up and running. Now was the time to link these two pieces

Linking Database to Application

In AppHarbor this linkage is achieved through Configuration Variables (CV). CV are variable defined for each application in AppHarbor. Once defined the variable names are matched with appsettings key names in configuration file (web.config) and value replacement done during deployment.

The following Configuration Variables are created automatically when SQL Server instance is provisioned

Variable SQLSERVER_CONNECTION_STRING_ALIAS is of interest to us. The value of this variable determines which connectionstring (in our config file) would be replaced by the value of variable SQLSERVER_CONNECTION_STRING

Nerddinner config file has two connection strings 'NerdDinners' and 'ApplicationServices'. Since I had just one database and just one variable replacement available I had to ignore one of the connection string ('ApplicationServices' in this case).

I achieved this using Config transformations. Since AppHarbor builds the solution in Release mode (currently no other mode supported) and also supports configuration transformations. I updated web.release.config with the correct transformations.

On AppHarbor front I had to set the SQLSERVER_CONNECTION_STRING_ALIAS value to ‘NerdDinners’ to achieve the desired linkage. The above variable is not directly editable from the Configuration Variables interface and requires one to dwell into the SQL Server add-on configuration.

Here is how this is done

Go to the application dashboard and click “SQL Server” add-on link

Click ‘Go to SQL Server’ link on the subsequent page.

This brings up the details for the provisioned SQL Server. Here we can change the alias

Once I had changed my alias, I pushed my config changes to remote repository and automatic build process kicked in. The site was updated within a matter of seconds and I had a fully functional Nerddinner up and running.

The site is available here http://nerddinner-4.apphb.com/ The code base is available here https://github.com/chandermani/NerdDinner70027

Final Thought

I recently deployed Nerddinner on IronFoundry, and now on AppHarbor. If you look at developer friendliness, AppHarbor wins hands down. The integration with code repository, easy\fast deployment, add-on services are some of the highlights of AppHarbor. Its deployment model is much superior to Windows Azure and Azure team can take some learning out of AppHarbor experience.

AppHarbor provides its infrastructure on per month rather than per hour billing. The same is true for AppHarbor add-ons too.

Update: As Rune commented 'AppHarbor actually prorates the monthly infrastructure and add-on prices down to the millisecond, so you only pay for the time you use it (even more fine-grained than EC2) '

IronFoundry still is in its nascent stage and a direct comparison would be unfair. Since IronFoundry is open source there is an opportunity to incorporate the good parts of AppHarbor and Windows Azure into IronFoundry platform.