Is it now OK to do piracy in Slovenia?

On 20th of October 2006 the Slovene government imposed an act of self reproduction compensation reproduction for personal use. Actually I don't know the proper English term so I'll describe what it is about.

The government wants us to pay some sort of compensation (tax) for every amount of storage we buy to compensate the damage to the authors done by piracy. This tax is applicable for all sorts of media: CD, DVD, tapes(what's that?) and other. There is more, even devices are included: hard drives, CD/DVD drives, tape recorders, video recorders, photocopy machines, mobile phones(!!!) etc. The tax is valued as €0.033 (or $0.04) per gigabyte of storage (and by some other mean for devices that don't have storage).

I am sure that's beautiful for the authors (or for the greedy government?). However, what does it mean for an user

  1. that uses licensed software only: Great, I am just buying a new computer with 4x320Gb hard drives and a DVD drive. That roughly means 4*€320*0.033 + €0.8 =~ €43. I am using licensed software only and paying for a piracy compensation. Does this mean I can use pirated software now? I paid for its compensation afterall.
  2. that uses non-licensed software: User pays for piracy now - is the piracy legitimate now?

Every (stupid - isn't it a requirement for politicians?) politician here has full mouth of e-[everything]. According to them we are going to be the most advanced/developed e-country by 2023 or something, yet they can't care less in reality. I strongly doubt that they even understand what e- means - it sounds nice and attracts crowds though.

So, while they have full e-mouths they keep adding taxes to devices and storage media. Aren't they supposed to lower the tax pressure and as consequence the prices instead of adding taxes to legitimate users?

What is the next step? We could get a stupid politician damage compensation: we would pay a tax when we vote to compensate the citizens for the damage done by elected politicians.

Anyway, since I am an author of my applications I'll apply for compensation...

Using client certificates with RemObjects' WinInetHttpClientChannel

Currently [RO]' Http channels don't support client certificates (i.e. client certificates are required in IIS when the option "Require Client Certificates" is selected). Why would you need them anyway? Client certificates are an excellent way of authenticating user and thus restricting the access to your website.

According to Marc, Chief Architect/[RO], client certificates support is planned for RemObjects SDK version 5 - next version.

But what if you need them now? Required change is really not that big but unfortunately isn't easily applicable. So [RO] guys suggested modifying WinInetHttpClientChannel.IntDispatch method where this feature belongs. It is certainly feasible and easy to apply, however I am not that keen to recompiling source codes. Not that it is difficult but when new version arrives you need to re-apply and re-compile.

That's why I went with deriving a new channel out of WinInetHttpClientChannel (note: you'll still need source code). Since IntDispatch method is declared as virtual so it can be overriden. The first pass is to create a derived channel named RhHttpClientChannel, add a X509Certificate2 field to it - we'll keep our client certificate in that field and a constructor that takes a X509Certificate2 as an argument. One more thing is to do done at this point. Since we'll be addressing a private field WinInetHttpClientChannel.fHttpHeader we need some reflection. The reference to this field will be stored in fHttpHeaderPI field and initialized in constructor. Here is the code so far:

public class RhHttpClientChannel: WinInetHttpClientChannel { private X509Certificate2 cert; private FieldInfo fHttpHeaderPI; public RhHttpClientChannel(X509Certificate2 cert) { this.cert = cert; fHttpHeaderPI = typeof(WinInetHttpClientChannel).GetField("fHttpHeader", BindingFlags.NonPublic | BindingFlags.Instance); }

Next, property HttpHeader, which is a wrapper to fHttpHeader, is implemented. Again, with a help of reflection (note that FullTrust will be required since we use reflection). fHttpHeader is a HttpHeader class. The problem is that HttpHeader class is declared as an internal in [RO] SDK assembly and thus can't be directly used. Lucky for us, it derives from Hashtable and we need only its Hashtable properties. So it is gonna be treated as a Hashtable.

private Hashtable HttpHeader { get { Hashtable value = (Hashtable)fHttpHeaderPI.GetValue(this); return value; } set { fHttpHeaderPI.SetValue(this, value); } }

Now you have to open WinInetHttpClientChannel source code and copy the entire method above (I won't show the implementation since I can't for legal issues)

private void CopyFromStreamToStream(Stream iSourceStream, Stream iDestinationStream, TransferDirection aDirection) { ... }

method (it is used by IntDispatch and private and I didn't want to bother with reflection).

Next step is to copy and paste IntDispatch method:

protected override void IntDispatch(Stream aRequestStream, IMessage aResponseMessage) { ... }

Inside the method fUserAgent with UserAgent and fKeepAlive with KeepAlive have to be replaced.

And the final step is to use the certificate that is passed to the constructor. Add this piece of code right after HttpWebRequest instance creation (first line below):

// original code lWebRequest = (HttpWebRequest)WebRequest.Create(TargetUrl); // add this below if (cert != null) lWebRequest.ClientCertificates.Add(cert);

There you go. You have a WinInetHttpClientChannel capable of using client certificates.

Transferring .net object graphs using RemObjects SDK 4.0

If you happen to use RemObjects SDK 4.0 (RO) for an alternative to remoting (I'll discuss the benefits in another post) than you are certainly facing one problem. RO thing doesn't support .net classes as method arguments in favor of interoperability (you can't use .net object graph from a different world, i.e. Delphi, right) - out of the box it supports only common value types (ints, strings, etc.) with some degree of extensibility. This isn't good when you want to push a .net object graph over the wire of course. Transferring .net object graphs works only in .net environment and it is not interoperable with other worlds. So as long as you are doing .net - .net you are fine.

The best option for this scenario is RO's Binary data type, a class derived from MemoryStream in version. I guess you see where I am going - I'll be doing manual (de)serialization to/from Binary.

Imagine we have a RO service method defined like this:

public Binary DoSomething(Binary args) { ... }

and in reality it should look like this:

public ReturnClass DoSomething(SomeClass args) { ... }

You will have to do some classic serialization:

public ReturnClass DoSomething(SomeClass args) { Binary binaryIn = new Binary(); BinaryFormatter formatter = new BinaryFormatter(); formatter.Serialize(binaryIn, data); binaryIn.Position = 0; Binary result = DoSomething(binaryIn);

and some classic deserialization:

result.Position = 0; BinaryFormatter formatter = new BinaryFormatter(); return (ReturnClass)formatter.Deserialize(result); }

However, when you deal with many methods that require this exercise the task soon becomes annoying. That's why I've created ROHelper class that does much of the (de)serialization for you. With some help of generics here it is:


public class ROHelper { public static T ConvertFromBinary<T>(Binary binary) { binary.Position = 0; BinaryFormatter formatter = new BinaryFormatter(); return (T)formatter.Deserialize(binary); } public static Binary ConvertFrom<T>(T data) { Binary binary = new Binary(); BinaryFormatter formatter = new BinaryFormatter(); formatter.Serialize(binary, data); binary.Position = 0; return binary; } }

and here is the DoSomething method rewritten:

public ReturnClass DoSomething(SomeClass args) { return ROHelper.ConvertFromBinary<ResultClass>( DoSomething( ROHelper.ConvertFrom(args) )); }

Isn't this code much better than spaghetti code above? Also note, that method ConvertFrom<T> doesn't require explicit definition of type T as it is deduced from argument itself.

The only thing that remains to improve is code generation itself. Even with simpler code like this you have to write all those method wrappers. So, the next step is to build code wrapper generator. Next time, perhaps.

Strongsigning an assembly from key container isn't supported even in VS2005 SP1

How can you strong sign an assembly using key located in key container (not in a file) from within Visual Studio 2005 (SP1)? You can't. For whatever reason MS decided not to give us this option in the 2005 RTM timeframe (apparently they didn't have time). Now, much later, they apparently still don't have enough time to implement it.

Why would one bother? Because storing strong key in key container (which is protected by OS) is more secure. It is also easier to reference such strong key as you need only its name and not the entire file path.

At least there has been a workaround which still works.

Windows Presentation Foundation PowerPoint slides available (Slovene)

Miha wearing medieval clothesYesterday I presented Windows Presentation Foundation. I was speaking in front of audience of about 150.  The whole event was using Columbus discovers America theme and we speakers were dressed accordingly - I suppose it was like a carnival.

If I take aside the funny outlook I have to say that it was a bit hot wearing additional medieval clothes. Anyway, the presentation was a bit overtime (ouch) because there are so many (good) things to say about Avalon WPF and there is never enough time just to explain the basics. I hope everybody had a good time.

Here is link to the slides (Slovene) and on the right is a photo taken with Qtek2020 PocketPC.

Can you spot the problem (VBx/VB.NET)


Here is a very popular problem that many VB developers fail to see. Furthermore, it is hard to find once in the code. Can you spot it?


imports Microsoft.VisualBasic imports System public module MyModule sub Main dim a as integer = 0 dim b as integer = CInt(iif(a=0, 100, SomeCalc(a))) end sub private function SomeCalc(a as integer) as integer if a = 0 then throw new ArgumentNullException("a") else return CInt(100 / a) end if end function end module

Why Microsoft keeps pushing laptop navigation software?

Microsoft has a range of GPS navigation products, such as Street & Trips, Map Point and Pocket Streets (did I forgot any?). The problem is that the former two run only on full blown Windows only (forget Pocket PC devices) while the later has no turn by turn navigation which makes all of three useless for in-car navigation.

I mean who would want to put a laptop on the windshield to have a navigation. This is so impractical that makes it useless. Even if you have a passenger that holds your laptop is really not usable.

On the other hand having a software (Pocket Streets) that runs on PocketPC and only shows your position on the map (the map that is pain to transfer from desktop PC to PocketPC) without turn-by-turn navigation is useless, too.

Combined with lack of map coverage for Slovenia makes it so useless combination that it almost hurts. I am struggling to understand the Microsoft strategy on GPS navigation software but I fail to see one. Even if I had Slovene maps it would be still useless.

Here is a free advice for MS: make a solid PocketPC (or Windows CE) turn-by-turn navigation software like the competition does (TomTom, Garmin, etc.) if you really want to have a car navigation system.

Importing Access database into Sql Server 2005

Today I had a need to import an old Access database into Sql Server 2005. First I tried with Sql Server Manager Import function but I quickly gave up on it as it was yielding many errors for some reason. Next I googled for a solution and found that MS is giving SQL Server Migration Assistant for Access (there are versions for other source databases) which is a full blown application that lets you configure many import parameters, mappings and other useful stuff.

It worked like a charm for me.