> (It also strikes me that .cvsignore files from upstream sources
  > probably should be pruned before import, along with files that
  > shouldn't have been in the upstream sources, but I know I'm ignorant
  > of the history of why things are as they are, so this may well be
  > wrong.)

  ".cvsignore" is the only entry in my "~/.cvsignore", so when I import
  a CVS checkout into my private CVS, the ".cvsignore" files are ignored,
  but also everything listed in said ".cvsignore" files is ignored.  Is
  that what you mean?

I meant two distinct things, but didn't discuss mechanisms.  Let me
try again.

  When third-party sources are imported, .cvsignore files should not
  be imported. [Mechanisms for this should be robust (which I think
  means not depending on cvs setup of individuals doing imports), so
  perhaps a per-commit check script and pruning is in order.]

  When third-party sources are imported, most files that the third
  party says to ignore probably shouldn't be imported.  If we take a
  'cvs export' of the third-party code, then the presence of any such
  files is likely a bug in the third-party code.  If we take a
  'distribution', then they might be files generated by tools that
  users are not expected to have (automake, gettext stuff, etc.).  In
  that case we probably are going to either not use those files, or
  run tools ourselves to generate them.  But, this may not always be a
  case, so I think judgement is required.

So I'm not sure automatically ignoring all things in the upstreams
.cvsignore files is right.  But it's probably close.

-- 
        Greg Troxel <gdt@ir.bbn.com>