[abacate]

The article is indeed interesting when it goes to the GC/allocation part and it's a nice read, but it really gets C++ wrong when talking about values/references. C++ is not "reference oriented language" the same way it is not an object-oriented language.

For instance, what restricts one from doing:

    int blah(myobj a);

Instead of:

    int blah(myobj &a);

For passing your a myobj value?

The usual argument is efficiency - since "myobj" may be a big and have an expensive copy constructor - and you may happen to see this pattern being commonly used in C++ programs, but restrictions? There are none. C++ even makes it possible to assert coherency with copy/assign constructors, and modern C++ supports the more efficient moving as well.

So, the point is: nothing in the language restricts you from passing things by value. I'd go further an say that blindly passing things by reference is bad practice and detrimental for simple objects - since you now have to consider lifetime/ownership of the reference you passed along, and this will probably complicate an otherwise simple piece of code.

The argument about fields in memory also makes not much sense. How is that different from C++? Keeping references/pointers in structures is possible but not exactly what one needs to do.

Otherwise the article is great I just think it misses the point when comparing to "C and C++", since they are not the same language - even though syntax is similar and there is some level of compatibility.