Messages in this thread | | | From | (Alan Cox) | Subject | Re: DHCPD v1.3 needs testing (fwd) | Date | Mon, 2 Feb 1998 19:18:28 +0000 (GMT) |
| |
> about providing a new API that can be used to better effect, but > shutting down the existing API on a moment's notice just isn't kosher.
There wasn't an existing API. I was _horrified_ when I found out how broken dhcpcd was in its way of working. And nobody changed an API, Alexey just added some IP standards compliance which broke an application doing what are basically broken things - although I can see why you did it that way.
> do away with the way Cox did. For years Cox insisted that BPF was > not useful and refused to put it in - that's why I support the API I > currently support.
BPF type stuff is in 2.1.x. It wasn't useful once. With 100Mbit cards and the bigger stuff Linux 2.x versus Linux 1.x is doing it now does make sense (actually it probably did for 2.0) - Jay wrote it and its in the kernel but done somewhat better in API than the original - you can push filters onto any socket - allowing stuff like user mode apps to do firewalling- it works so much better when your inetd can read the hosts.deny list and tell the kernel to block packets to this socket at entry not in userspace.
Someone has done a nice SOCK_PACKET based change to dhcpcd that seems to work well for 2.1.x and 2.0.x
> The parallels to Microsoft's way of doing business are positively > uncanny.
If I made you send me $500 for each time I break your program then maybe you'd have a vague point. I also don't charge $99 an email for answering queries or requests for fixes - although direct those to the current networking maintainers rather than me
Alan
| |