newbie question regarding "cvs init"

[Robert P. J. Day]

  hard to believe but i have a question regarding creating a new
repository with "cvs init".

  i'm aware from the docs that the standard practice is to create an empty
new CVS repository, and populate it from there with "cvs import".  
however, is it feasible (or just plain bad practice) to create a new
repository over top of an existing directory structure?

  that is, if i have a directory "dir", under which i have
several project subdirs, "s1", "s2" and so on, i can apparently
do something like

  $ cvs -d dir init

which will create the CVSROOT directory inside "dir", at which point
i can checkout any of s1, s2 and so on.  so, technically, it seems
doable, but is it just considered bad form to drop a repository over
an existing directory hierarchy?

  and, second, from all i've read, it seems i can only use "cvs import" to
import what's in my current working directory, is that right?  the options
list for "import" doesn't appear to accept arguments of what to import, 
and only takes the contents of the current working directory, correct?

rday