“C++ Coding Standards” by H. Sutter et al, Addison-Wesley, Feb 2005.

for two primary reasons:

1: Safety in code maintenance. Many times I have seen the alternate format accidentally split, changing the code flow but not generating compile warnings:
if ( x )
{
blah blah;
}
else
line added during maintenance;
{
blah blah;
}

or the simple case
if ( x )
line added during maintenance;
{
blah blah;
}

The most common instance is the accidental
for ( ix = 0; ix < END_IX; ix++ );
{
copy[ix] = original[ix];
}

Legal code and hard to spot, but not what is intended.

2: Compact presentation. Either on an old-fashioned glass teletype screen, or in a typical small window in a GUI environment, one is often performing keyhole surgery on the source code, and this format maximises the quantity of functional code visible.

I also like the way that the closing brace matches the start conditional (or function, or loop control, etc), and the clearer indentation (IMHO).

The convention

if ( x ) {
blah valh
}else {
}

is awful and I prefer a more compact manner like this:
if (x)
{
doSomething1();
doSomething2();
}
else
{
doSomething3();
doSomething4();
}

or

if (x)
doSomething1();
else
doSomething2();

But I appreciate the fact id Software has a coding convention paper and this is a good example to follow with of course, each one with his own conventions!

I’m programming C++ for 10 years and that doc clearly shows some immature decisions, for example:

if ( x ) {
blah valh
}else {
}

that’s ugly notation that does not help at all the code readability and neither correct code identation. It’s way better this other one:

if ( x )
{
//blah blah
}
else
{
//blah blah
}

( which is called ANSI C++ style brackets btw ).

Other thing I don’t like is that pseudo-hungary notation with the _t at the end

struct myStruct_t

almost all the code editors in the world ( VS, NetBeans, Eclipse, SlickEdit, etc etc ) can tell you if a data structure is a struct, a class, an integer, etc in the debug watch, class navigators, etc… actually, I think there is no need at all to use hungary notation and neither that 80-era _t thing.

They’s are using “float” also which is terribly bad. A good practise is to re-map floating point types to custom typedefs or classes, so you can alter the type them easy. For example, imagine you’re developing a mobile game and that platform does not suppport floats… with custom classes or a typedef you can use fixed fp precision or whatever…