I use spaces in page names keeping lower and upper case as per common rules for titles. A also use diacritics, punctuation etc. I use UTF-8 exclusively. I do not care how it is encoded in the background (if even needed today) for my browser can handle it properly for me. I would prefare to have everythink as it is on the filesystem as well to navigate normally from the command line for my shell handles it correctly for me too. (recent Linux, Firefox, BASH)
There is the filesystem end and there is the browser end. SilverBullet should just “proxy” between HTTP URL encoding back and forth on both ends (if needed). If I name a page Příliš žluťoučký kůň! I should be able to write it as [[Příliš žluťoučký kůň!]] as well as [Příliš žluťoučký kůň!](/Příliš žluťoučký kůň!) and so will be named on filesystem as well. As user I must not be forced to handle or work arround the need for the URL encoding because of some historical flaws in some protocols. My user naming schema should not be disturbed in any way. I think this is crutial.
If it means special handling for some OSes or anything like that, let’s do the extra work. The benefit of having this right and seamless is of the highest importance!