OpenBSD's traffic shaping changing in future releases

This initial change is relatively minor but will likely give a performance boost right away. I imagine it will also help introduce basic shaping in a lot of configs via per rule changes like the ones quoted in the article and messages.

(It has been a while since I've gotten armpit deep into a gnarly pf config file, so I might get part of the following a little bit wrong. I welcome corrections.)

pf has two different packet shaping mechanisms: CBQ, and prio. CBQ allowed you to allocate amounts of bandwidth, but didn't include prioritization. So, you could, say, set aside X Mb/s for ssh, Y for http, and so on, with lots of borrows and congestion management and other neat stuff.

With prio, you had to give up allocation, but you got to say instead that a particular rule always got bandwidth priority over other rules. So, if your web server is getting hammered, and you have ssh prio 7, you can still get into your machine no problem.

To use either scheduler, you had to specifically "turn it on" in your config file and essentially build a little structure for it. It sounds like -- having just glanced at Henning's messages and patches -- the "turn it on" and structuring steps have been eliminated for prio. Although on the surface that looks like a small change (and user-wise it probably is), it's still a pretty big change in the pf internals.

What I hope -- and can't tell yet -- is that this is going to get us a step closer to being able to use both scheduling mechanisms. They are complementary systems and it sucked a little to have to choose the features from only one or the other.

Either way, this is the kind of news that makes me really really miss working with pf. I prefer it, a lot, over iptables, which makes me grouchy on a regular basis.