On Fri, Dec 29, 2000 at 05:40:20AM +0000, Charles M. Hannum wrote:

 > * First of all, I agree with the concept.  (After all, I proposed the
 >   idea years ago myself.)  However, it doesn't make sense to ask
 >   people to give you a green light to commit something based on a man
 >   page alone.  Even disregarding actual code, there are many things
 >   you've left unspecified -- such as where it actually hooks into the
 >   network stack.

Ah, you're right.  It uses the PFIL_HOOKS mechanism.  So, take a look
at where those hooks are currently.

For network interfaces, which I have not committed yet, each network
interface has a pfil_head, and e.g. ether_input(), fddi_input(), etc.
would be responsible for running the hooks if there are any enabled
for that interface.

 > * I have SERIOUS reservations about NetBSD components being named
 >   after companies.  This sets a bad precedent, and I envision a day
 >   when it won't be possible to even read documentation without seeing
 >   what is effectively advertising.  This was never done in the
 >   Berkeley days, either, even though many components (e.g. label
 >   support) were contributed wholesale.  NetBSD is an operating system,
 >   not an advertising medium.

Heh, amusingly, the "Zembu Packet Classifier" name is pretty new (it
was called something else that didn't have "Zembu" in the name, but
was renamed after IETF after people commented that "... filter" wasn't
a descriptive-enough name, considering its capabilities).

Renaming it again is easy (tho annoying).

-- 
        -- Jason R. Thorpe <thorpej@zembu.com>