Implementing the Win32 API provides a particular challenge to anyone hoping to get involved in this area, in that it has to mimic Windows versions of the Win32 API completely. This means not only implementing the features so they operate exactly as they do on Windows, but also implementing their bugs too. Implementing bugs sounds strange but some applications rely on unintended outcomes of particular APIs, and unless they introduce security issues, then we need to make sure our code has these quirks too.
I think thats rather curious in some way.
I of course concur that the api should run everything that is possible to run on windows, however those bugs - if they really are bugs - should not be the target of the daily work platform, i would think to the normal os running on all bugs fixed but having a compatibility mode for applications that rely on the quirks.
That's sort of how I see it. If they truly are bugs, then fix them and provide a compatibility option. But if they are simply quirks or undocumented features, then by all means keep them as close to Windows as possible.
Unfortunately, I had trouble with Archive.org's Wayback Machine when I wrote my earlier post that mentioned the now-obsolete link. My usual practice is to archive any links I mention in my comments here, if the site allows it. Try as I might, I could not get the Wayback Machine to archive that page. So the very useful and informative comments from 2003 were lost. Fortunately, Raymond Chen worked hard to make sure that his old links are redirected to the corresponding URIs at his new blog so his oldest posts can still be found by following the old links.