> Second, on readability and comprehensibility, I find two spaces is
> simply too little. Try reading glibc code sometime; two spaces is
> just not enough.  _If_ we decide to change, and M$ data says four
> spaces are as legible as two, I put in a strong plea for four.

So, I have a couple of cents to throw in.

First, when I started working for my group at MIT, most of the
programmers in the group preferred the gnu indentation style, so we
standardized on that.  I thought I would hate it (being a four-space
K&R-style man by nature), but after a surprisingly short adjustment
period I found that it was just as readable as what I was used to.

So it may be that the studies Mike Cheponis cited weren't much more
than a popularity contest; if most of the participants were used to
2-4 space indents, then perhaps they'd be able to read them better.
Not having a copy of the original article, I can't be a good judge.

Second, I'll claim that two-space indents make more sense when you're
putting braces at their own level of indentation.  The gnu indentation
style looks like:

	if (foo)
	  bar;
	if (baz)
	  {
	    quux;
	    quuux;
	  }

so complex bodies are actually indented at four spaces.  (Obviously,
some people hate the fact that simple bodies are indented at a
different level than complex bodies; I refer them back to my first
point.)

Third, changing the indentation style of the KNF would be awfully
traumatic.  You could say goodbye to automatically applying any
patches from non-reindented code.  Since by all indications
indentation style is a fairly arbitrary choice, I think changing it
would be a big waste of effort.