Hi, Shiro.  Thank you for this pull request.  I haven't heard from Will Clinger, to whom I wrote privately about these changes.  However, it's not necessary for the author to approve changes to the sample implementation, so I just reviewed them myself.

The changes look good except for one thing: the value of assumed-endianness is never assigned, so the tests won't actually work in the little-endian case.  It looks like #f in (define assumed-endianness #f) should be replaced with the six-line expression that immediately follows it.

If I'm right, would you please make that change to the pull request?  I'll publish it immediately afterwards.

Thanks.