Yes, I see the smiley, but I thought it would be fun to measure this effect anyway. I whipped up a small Python script to measure the potential savings, and ran it across the 450kLOC GlusterFS codebase. It reported 3.4MB of savings (both realized and potential because our code isn't as consistent as it could be). I'm generally in the spaces camp myself, and believe that 3.4MB just isn't worth worrying about any more (despite being worth ten of the floppies that were omnipresent even several years into my career), but it's still a surprisingly high number.