On Idiot Proofing and Future Proofing

You can attempt to idiot proof something by guarding against mistakes typically made by the idiots of today.

You can attempt to future proof something by guarding against foreseeable mistakes that might be made by reasonable people at some point in the future.

However, you cannot future idiot proof anything. Idiots keep getting more idiotic. There is no way to predict what mistakes the idiots of tomorrow might make.

I curse you.

This is my message to the following people.

  • Anyone who is anti-LGBTQ+.
  • Anyone who is anti-immigrant.
  • Anyone who is anti-abortion.
  • So many others.

I curse you. I cast upon you the greatest curse I have the power to cast.

May you live a long and pointless life. And many years from now when you are laying on your death bed, may you find yourself alone instead of being surrounded by friends, family, and loved ones because you have driven them all away through your evil, hateful acts. And may you, on the day you die, finally realize and accept just how evil you are. And may your last thought in this life be as follows. “What awaits me? Will it be a reward? Will it be rest, peace, and forgiveness? Or will it be a well deserved punishment?”

The Carbon Programming Language is not a legitimate successor to the C++ Programming Language.

The Carbon Programming Language was recently announced as an experimental successor to C++. In my opinion the Carbon Programming Language is not a legitimate successor to the C++ Programming Language.

If you create a new programming language that is meant to be a successor of an existing programming language, do not break every single rule of the original language.

In C, C++, C#, and many simular languages data types are on the left and variable names are on the right. Anyone who has any experience in these programming languages expects any programming language that claims to be a successor to follow this most basic of syntax rules. This rule is such a basic rule of the language you should not even consider breaking it for a so called successor language.

The Carbon Programming Language breaks that rule.

Maybe people who have not ever used C and C++ will be willing to learn the Carbon Programming Language, but people with 20+ years of experience will hesitate because the code is too different.

This should be obvious.

Putting the variable name on the left and the data type on the right may make parsing easier, but millions of people who have experience in C or C++ or C# expect the data type to be on the left. Google, by going against these expectations is greatly limiting the adoption of the Carbon Programming Language.

Why does Visual Studio always do this?

Why does Visual Studio always do the wrong thing when creating a new solution or project file in an existing directory.

Consider the following scenario.

  1. There is source code but no project file in the directory c:\Path\To\ProjectName.
  2. You create a new project file named ProjectName in the directory c:\Path\To\ProjectName. You expect the project file to be directory c:\Path\To\ProjectName\ProjectName.vcxproj because that makes sense.

    Note that ProjectName is already repeated twice in that path.
  3. What Visual Studio actually does is create a directory named ProjectName in the directory c:\Path\To\ProjectName and then places the file ProjectName.vcxproj in that directory so that the full path to the project file becomes c:\Path\To\ProjectName\ProjectName\ProjectName.vcxproj.

    Now you have a path where ProjectName is repeated three times.

Every time this happens I have close the solution file, to go to the directory in Windows Explorer, move the project file up a level, delete the extra ProjectName directory, manually edit the solution file in a text editor, then reopen the solution file in Visual Studio. Please, just put the project file in the specified directory.

Friendly Reminder: If you release a NuGet package please sign all the DLLs.

Please. I am begging you. If you release NuGet packages sign both the Debug and the Release DLLs. Do not just sign the Release DLLs.

I wasted four hours today debugging the following exception: “FileLoadException: A strongly-named assembly is required. (0x80131044)”.

I was debugging the debug build of an application that uses the Titanium.Web.Proxy version 3.1.1397, obtained using NuGet. The Titanium.Web.Proxy uses BouncyCastle.NetCore version 1.8.10.

I have a python script that copies the build output from the output directory to the appropriate location on my test PC where I have the product I am working on installed. This python script replaces the files already present in the application installation directory with the files produced by building the product on my development PC.

The build process, which uses MSBuild, copies the DLLs for BouncyCastle.NetCore and Titanium.Web.Proxy to the output directory. This is useful for release builds because the installer project pulls all DLL and EXE files from the build output directory. My python script previously copied the BouncyCastle.NetCore and Titanium.Web.Proxy files to my test PC in addition to all the DLL and EXE files for the application.

The problem is that the Debug DLLs for BouncyCastle.NetCore and Titanium.Web.Proxy are not signed. When a .NET assembly depends on another assembly and the dependent assembly is not signed .NET will throw a “FileLoadException: A strongly-named assembly is required. (0x80131044)” exception when the first assembly attempts to load the dependent assembly.

It took me four hours to find out why this exception was occurring.

Advice on writing a clear and correct technical article.

Woman screaming in rage,

Please, if you are writing an article like Git Fixes: The Current Branch Has No Upstream Branch in which you attempt to explain how to solve the “fatal: The current branch has no upstream branch” error, do not use the wrong branch name in your sample commands.

For example, do not claim that the problem can be solved with the following command.

git push -u origin master

This error cannot in any way be solved by the above command. Why? Because as you correctly point out in the beginning of your article, the reader is not working in the master branch. The only reason the reader saw the error in the first place and did a Google search to find your article is they created a local branch that is most definitely not named master and an error occurred when they attempted to push the changes. The above command is only valid if they are working in the master branch.

The moment the reader sees master in the above command when they know for a fact that it is not correct the begin to question you credibility.

You started off so well in the third paragraph when you included the following.

git push --set-upstream origin <branchname>

In the above example <branchname> makes it clear that the reader is supposed to insert the name of their own branch in that location in the command.

You started off so well. But then you messed up by putting master where you should have put <branchname> throughout the rest of the article. The moment a reader sees master where they know for a fact that it does not belong, your poor reader may go to another less detailed source rather than figure out what you meant to say. I had to read ten other less well written articles before I figured out what you meant to say.

Reading your article could have been a pleasurable experience causing me to be thankful that the solution was so easily found and implemented. Instead it became an hour long headache.

Computer manufacturers and computer accessory manufacturers really need to use consistent terminology.

Frustrated woman. From pxfuel.com.

Computer manufacturers and computer accessory manufacturers really need to use consistent terminology. Their failure to do so makes things unnecessarily difficult.

I am considering buying a computer from Dell since one of my main computers does not support Windows 11. According to Dell the computer has the following display ports: 1x HDMI 2.0b, 1x DisplayPort 1.4, 1x Dual Link DVI.

The problem is that I need to be able to use this computer with an existing KVM that uses VGA. Therefore I will need to use adapters. The question is what adapters do I need.

Computer accessory manufacturers do not use the Dual Link DVI terminology at all. They use DVI-D and DVI-I. In order to determine what adapter to get I had to go to the Different types of monitor ports page and compare the pictures on Dell.com with the pictures on the Different types of monitor ports page to determine that Dual Link DVI is apparently the same as DVI-D. Therefore the Active DVI-D to VGA Adapter should allow me to connect a VGA monitor cable to the Dual Link DVI port.

Computer accessory manufactures do use the term HDMI. The problem is that not one of the search results for HDMI 2.0b to VGA adapter say anything about the HDMI version. They are HDMI to VGA adapters. No version number information is provided at all. I have to hope that the Gold-Plated HDMI to VGA Adapter will do the job.

Thanks to the use of inconsistent terminology buying this computer has been an arduous and very unpleasant experience.

Thank you very much!


User Interface Developers: Please Have Mercy on Your Poor Users!

Woman screaming in rage,

This blog post is for any software engineers who are involved in programming or designing a user interface.

If someone performs a user interface action that should result in an instantaneous and clearly visible response they will only wait patiently for that response for a brief period of time.

If after a five second delay they do not see the expected clearly visible response they will perform the user interface action again, perhaps multiple times if the delay continues for another five seconds or so.

If for some reason there is a delay that prevents the expected instantaneous and clearly visible response, please, for the love of whatever deity or great power you follow, DO NOT PERFORM THE RESPONSE MULTIPLE TIMES.

The multiple user interface actions were not in any way a request to perform the response multiple times. They were multiple increasingly desperate attempts to get the response a single time.

For example, if the user is pressing enter on a desktop icon that starts a program that displays a message box before the main function is even called, they will press enter on the desktop icon multiple times if for some reason it takes between five and sixty seconds to start the application when it normally starts in far less than a second.

Trust me. The user has already started screaming at you by the time they have to hit enter on the desktop icon a third time because the message box still has not been displayed in response to the request to start the application that was made fifteen seconds earlier.

If you then start that application one time for each time they pressed the enter key on the desktop icon when your badly broken operating system finally realizes the user is desperately trying to start it the poor user might end up having a stroke because of the rage they are now experiencing!

Another example is opening a menu such as the start menu or an application menu. Normally this is a very quick operation that takes less than a second. So if five seconds after they pressed the key to open the menu the user does not see the menu, they will press the key to open it again. On the rare occasions when something delays the opening of the menu they may have pressed the open menu key four or five times before it is opened.

That does not mean they want the menu to be opened, then closed, then opened, then closed…

The multiple presses of the key to open the menu were multiple increasingly desperate attempts to open the menu once. By the second attempt the poor user is screaming at you. If you do the open menu, close menu, open menu, close menu, … foolishness it is very likely that the poor user might end up have a stroke because of the rage they are now experiencing!

These are problems I experience very often with Microsoft Windows. To be fair I have seen it with Linux and Mac OSX as well, but no where near as often as I see in in Windows.

Fuck you very much Microsoft for the wonderful user experience of Windows and every single Microsoft product I have every used! If I believed in hell I would wish for everyone responsible for these wonderful user experiences to suffer eternal torment in the hottest and most fiery section of hell!

Thank you for listening to me vent. I feel much better now.

Here is a song that expresses my heartfelt feelings towards Microsoft. Enjoy.

… you very much Microsoft

On the benefits of taking a course over learning by doing.

I am a self taught software engineer. It is something I take a great deal of pride in.

I started learning C programming language back in 1997. If I remember correctly I started by reading Sams Teach Yourself C in 21 Days and perhaps also C Programming Language by Brian W. Kernighan and Dennis M. Ritchie.

Then on January 5, 1999 I got a job working for Henter-Joyce (now Freedom Scientific) as an Associate Software Engineer. My responsibilities included writing code in the JAWS Scripting Language which a proprietary programming language that is used by the JAWS screen reader.

I quickly decided that I was not content with just writing code in the JAWS Scripting Language; I wanted to be able to modify the JAWS internals.

At that time almost all of the JAWS internal code was in the C++ programming language. Therefore I decided to teach myself C++.

To learn C++ I did the following.

  • I bought a copy of Sams Teach Yourself C++ in 21 Days and began reading it.
  • I wrote an application called SysInfo in C++ as a aid to our technical support department. SysInfo would gather information such as Video Card information, OS, memory, and send it to technical support via email. SysInfo was very similar to the current Microsoft System Information Tool.

This plan was so successful that I was promoted to Software Engineer in January 2000. I have used C++ as my primary programming language since then.

I never stopped learning. I have tried to keep up to date with new language standards and new technologies since then. But I always learned by doing. If the task I was working on required that I learn something new I would use Google and Stack Overflow and get the job done.

This has been my preferred method of gaining new knowledge until very recently.

On August 1, 2021 my knowledge of JavaScript was minimal. The last time I needed to use JavaScript as part of my job it took me two weeks to fix a bug that required that I change fewer than 10 lines of code.

Then I realized that there are some serious flaws with a browser extension for Google Chrome and Mozilla Firefox released by my employer that I am currently responsible for maintaining that necessitated a complete rewrite of the extension. I tried to make the necessary changes and found much to my horror that I could not even get started.

I had previously bought a lifetime membership to Eduonix and some courses on Udemy with the vague idea of taking the courses some day. However, I had made very minimal use of these educational resources.

I let my employer know what was going on and I was given permission to spend some time when I would normally be working on taking courses in JavaScript. I then spent every day for the next two weeks learning JavaScript.

In just four hours this Monday I rewrote one quarter of the Browser Extension. There is only one reason I did not finish the project; I was temporarily reassigned to another project.

I am not claiming that my knowledge of JavaScript is anywhere near my knowledge of C++. However, after just two weeks my knowledge of JavaScript is roughly equivalent to my knowledge of C++ back in 2001.

Looking back, I find myself wishing that I had followed a more structured approach to learning C++ back in the halcyon days of yore.


JAWS Scripting Tip: Optional Parameters

JAWS is a screen reader developed by Freedom Scientific. It includes a scripting language that allows users to customize the behavior of JAWS.

Many programming languages such as C++, C#, and Modern JavaScript support the concept of optional function parameters for which you supply a default value for the parameter. In C++ this might be done as follows.

std::wstring GetFavoriteNumberMessage(int64_t number = 42)
    std::wstringstream message;
        << L"Hello. My favorite number is "
        << number
        << L". What is your favorite number?\n";
    return message.str();

See source on godbolt.org.

The JAWS scripting language supports optional parameters using the optional keyword, as follows.

string function GetFavoriteNumberMessage(optional int number)
  var string message = FormatString(
    "Hello. My favorite number is %1. What is your favorite number?",
  return message

However, there is no direct method of assigning a default value for the optional parameter. If you were to call the function is as without specifying a value for the number parameter the number will be inserted into the returned message as an empty string. This is most likely not the desired behavior.

When the caller does not specify the value of an optional parameter, it does not resolve to 0 or false within the function; this is a common misconception. Instead the unspecified parameter resolves to nullptr. The misconception that the unspecified parameter resolves to 0 or false is a result of the fact that when nullptr is typecast to an int the result is 0 or false. However, if the unspecified parameter is typecast to a string the result is an empty string. This is why in the above example the number will be inserted into the returned message as an empty string; the number parameter is typecast to a string by the FormatString function. If the unspecified parameter did resolve to 0 or false within the function it would be inserted to the returned message as 0!

This problem can be resolved as follows.

string function GetFavoriteNumberMessage(optional int number)
    if (!DidCallerSpecifyTheValueOfTheParameter(number))
      number = 42
    var string message = FormatString(
      "Hello. My favorite number is %1. What is your favorite number?",
    return message

Unfortunately, there is no DidCallerSpecifyTheValueOfTheParameter function in the JAWS scripting language, so do not bother looking it up. However, there is a function that can achieve the same goal.

Behind the scenes all variables in the JAWS scripting language are VARIANTs (yes we are aware that there are modern alternatives that are vastly superior but the JAWS scripting language is very old; it has its roots in the JAWS macro language that was included in JAWS for DOS). One benefit of a VARIANT is it supports the concept of a VARTYPE or variant type. Furthermore, there is a VT_EMPTY variant type. Finally, there is a rarely used GetVariantType built-in script function that “determines the type of a variable”.

Thus, it is possible to assign a default value for an optional parameter as follows.

string function GetFavoriteNumberMessage(optional int number)
    if (GetVariantType(number) == VT_EMPTY)
      number = 42
    var string message = FormatString(
      "Hello. My favorite number is %1. What is your favorite number?",
    return message

Note that some people may think this is unnecessary because it is enough to rely on the perceived unspecified parameter resolves to 0 or false behavior. This is wrong for two reasons.

  • As I have already pointed this perceived behavior is not what is actually happening. The unspecified parameter actually resolves to nullptr.
  • In many cases, even if the perceived unspecified parameter resolves to 0 or false behavior may not be what is desired.

Every other programming language that supports the concept of optional parameters; such as C++, C#, and Modern JavaScript ; allows the value of the optional parameters to be explicitly defined. The purpose of the following code block is to add the same capabilities to optional parameters in the JAWS Scripting Language.

if (GetVariantType(number) == VT_EMPTY)
  number = 42

This is the only way to assign a non nullptr value for an optional parameter in the JAWS Scripting Language. You should consider doing this even for parameters where the default value is 0 or nullptr because explicitly specifying the value of your default parameters rather that making people who read your code guess what the behavior will be is just being polite.

Note that using this technique is absolutely necessary when overriding a built-in script function such as the SayObjectTypeAndText function in your scripts. The default value of the includeContainerName is true, not false. Failure to do this when overriding the SayObjectTypeAndText function will change the behavior of your function from the default resulting in hard to find bugs.