Why is the term "string" so often abbreviated as "sz"?

Its a variable naming convention called Hungarian Notation. It was common in the 90's, and notably used a lot by Microsoft in their Windows API docs. The idea is to prefix variables with hints about their type. The exact prefixes used varied, but you might see a variable containing a count called nCount for integers, fCount for floating point values, etc.

The sz stands for 'zero (null) terminated string'. Basically a character array that ends in a null. You'll also see 'lpsz' used, 'long pointer to a string'.

If your first reaction is to say why was this needed when you can just hover over a variable in your IDE and see its type, you have to understand that such IDEs did not always exist.

A lot of people still use the general idea of Hungarian Notation these days, but rarely as rigidly as it was used in the past.

As @GrandmasterB says in their answer, this is "Hungarian Notation".

The term "Hungarian" originates from Charles Simonyi, a computer programmer of Hungarian background, who worked at Xerox and later Microsoft. He conceived and popularised this particular system of name prefixing.

The sz prefix means a zero-terminated string.

In both of the two main character encoding schemes, ASCII (for PC) and EBCDIC (for mainframes), the zero value means the Null character, hence the alternative (and probably more common) term, null-terminated string.

Terminated strings and string arrays are common in the Windows API and in the C programming language more generally, and so there is a third common term for this, "C-style strings".

The main alternative to null-termination is called length-prefixing, also called "Pascal-style strings", because the Pascal language used this approach for its native concept of strings.

Length-prefixing is where a value is stored in memory before the string, indicating the length of the string. My understanding is that the Hungarian prefix for this type of string is st, although I've never myself encountered this prefix.

There are different pros and cons to each approach, but the tendency of modern languages is to employ length-prefixing of strings.

Then, in Microsoft's COM technology, you have the BSTR, which is a string that is both length-prefixed and null-terminated! The common Hungarian-style prefix for this was bstr.

The bottom line is that there are a lot of different approaches to encoding strings, depending on language and context, without a specific standard.

In terms of facilities to cope with such variety, languages like C are now very old, and it was not designed with a very sophisticated type system. The typing system is so basic in C that any kind of string is considered a "complex" data type cobbled together from its elements, so although C is capable of handling all kinds of string encoding, there aren't built-in types for any of them.

C was also designed in an era before modern IDEs and when source code was often printed on listing paper, so there are fewer alternatives for coping except to encode things in the text of the source code.

Under those circumstances, a system of naming prefixes is absolutely necessary to stay on top of which data types you are dealing with, if you're in an environment where one particular approach to strings cannot simply be assumed as universal.

The Windows API itself is now of a similar vintage as the C language, and Windows was originally written predominantly in x86 assembly language and in C.

And Simonyi worked at Microsoft, and therefore was in a position to influence how they coped with the programming challenges of the era - including the fact that programming at Microsoft consisted of more than just Windows and its conventions.

That is basically why the use of null-terminated strings, and the use of Hungarian Notation, is widespread there.