Hacker News Reader

Comment by deafcalculus | Hacker News Reader

Was delayed ACKs the problem? Disabling delayed ACKs seems like a better bet than using TCP_NODELAY which turns off Nagle's algorithm.

View on HN

I generally say yes. The fixed timer is for the delayed ACK. That was a terrible idea. Both Linux and Windows now have a way to turn delayed ACKs off, but they're still on by default.

TCP_QUICKACK, which turns off delayed ACKs, is in Linux, but the manual page is very confused about what it actually does. Apparently it turns itself off after a while. I wish someone would get that right. I'd disable delayed ACKs by default. It's hard to think of a case today where they're a significant win. As I've written in the past, delayed ACKs were a hack to make remote Telnet character echo work better.

A key point is asymmetry. If you're the one who's doing lots of little writes, you can either turn set TCP_NODELAY at your end, or turn off delayed ACKs at the other end. If you can. Things doing lots of little writes but not filling up the pipe, typically game clients, can't change the settings at the other end. So it became a standard practice to do what you could do at your end.

You should sign your post with your name (in this topic), in case a lot of people are not familiar with who you are :).

For others, their name is in their bio and is John Nagle.

Disagree. It's a lot more fun when everyone has to figure it out for themselves. :)

On macOS add the line net.inet.tcp.delayed_ack=0 to the file /etc/sysctl.conf (which doesn't initially exist).

Some linux settings apply to every TCP connection which can be a very surprising result and hard to debug. I can see someone wanting to time out a feature, so it doesn't upset the rest of the system forever.

The article is refreshingly short and to the point. Unusually, it does not aim the waste the maximum amount of your time so why don't you read it.

I liked the article, but it's not entirely clear to me what the cause of the problem was. Linux delays sending out data only if the data size is less than the packet size AND the previous packet in not yet ACKed (Nagle's algorithm). My guess was that this app is doing a write, write, read and hitting the delayed ACK problem.

The problem with delayed ACKs is that requires controlling the server. If you control the client, you can’t remotely turn off delayed ACKs, so instead you have to disable Nagle’s algorithm.

Reply on news.ycombinator.com