Comment 1 for bug 2062018

> The one performance improvement (that I didn't benchmark): "Make compression libraries dynamically loadable" (it's from January, not April).

Should apparently make the initialization quicker. Nice to see this coming from a reputable source (@redhat.com) ;)