On Thu, Mar 09, 2006, David A. Case wrote:
> From: Scott Brozell <sbrozell.scripps.edu>
> In 8 there was an undocumented pb namelist variable nsaa.
> In 9 there is an undocumented pb namelist variable maxarc.
This is a tough call, and I have not been consistent in the past. One the
one hand, having undocumented options is attractive to developers. But it
has turned out the users try them anyway, future developers are confused
by them, and those options often become broken by future code changes,
because
they are not being tested.
So the new policy is: no undocumented namelist variables. (Scott's
"retired"
varaibles are OK: they stop execution with an informative error message.)
If you need local variables for your personal work, make a CVS branch.
[As with everything, exceptions may be granted, but I think this is the
correct general rule. A corollary is: don't leave "commented out" or
"ifdef-ed out" stuff in the code. Such stuff is as likely to be confusing
as
helpful, and "turning it back on" may break things since the code has
evolved
since it was originally written.]
We'll try to get the above rules implemented as much as possible by the
time Amber 9 comes out.
...thx...dac
Received on Wed Apr 05 2006 - 23:49:40 PDT
