Re: Slow TCP and everything

Linus Torvalds (
Mon, 15 Apr 1996 08:27:08 +0300 (EET DST)

On Sun, 14 Apr 1996, Benjamin Lahaise wrote:
> The patch also corrects a few ntohs->htons's (DOH!), a potential problem
> in option parsing, and why weren't we throwing away duplicate packets?
> (Although thinking about it, it _might_ be redundant...)

I took out the code that throws away duplicate packets on purpose: it
only complicates the normal TCP path of control, and it's useless anyway
(as you suspected). Duplicate packets are later removed when we read the
TCP data, or in the special "prune_queue()" (or whatever I called it)
function that is called only if we notice that the receive queue has grown
too much (ie never, under normal circumstances).