On Sat, 10 Oct 1998, Stan Ryckman wrote:
At 05:18 PM 10/9/98 -0700, Bart Schaefer wrote:
As a first contribution to this effort, here's an extremely simple bug fix
for the "multigram" program to address a problem that was mentioned on the
"Ongoing development" thread:
On Thu, 8 Oct 1998, mark david mcCreary wrote:
A big problem is that it does not handle different level of domain
nodes correctly. That is, joe(_at_)email(_dot_)msn(_dot_)com does not
match up to
joe(_at_)msn(_dot_)com(_dot_) Very frustrating for both joe and the
listmaster.
That's not a bug... those are *DIFFERENT* email addresses. In some domains,
they do NOT correspond to the same user, or one may be his shell account
and the other his ppp acount.
Whether it's a good idea for multigram to score those addresses as similar
is open to debate, but the algorithm *is* intended to do so.
If you look carefully at the findatlast() function in multigram.c, you'll
see that without the patch I posted, it *always* returns NULL. That means
that an entire branch of lowcase() is never executed. That can't possibly
be what SRB intended.