<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>TCP on 菜菜博士</title>
    <link>https://microcai.org/tags/TCP.html</link>
    <description>Recent content in TCP on 菜菜博士</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Tue, 06 Oct 2015 00:00:00 +0800</lastBuildDate>
    <atom:link href="https://microcai.org/tags/TCP/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>UDP and congestion control</title>
      <link>https://microcai.org/2015/10/06/udp-and-congestion-control.html</link>
      <pubDate>Tue, 06 Oct 2015 00:00:00 +0800</pubDate>
      <guid>https://microcai.org/2015/10/06/udp-and-congestion-control.html</guid>
      <description>&lt;p&gt;TCP is good, good for nearly everything. It&amp;rsquo;s a general purpose abstraction for networking applications. But TCP is bad for one thing: it&amp;rsquo;s general. If you have some special need, setsockopt(2) can help you. But really, it dosen&amp;rsquo;t help much.&lt;/p&gt;
&lt;p&gt;What if you have customized need for congestion control?&lt;/p&gt;
&lt;p&gt;The need for different congestion control algorithm is the main reason that we implement
our protocol ontop of UDP.&lt;/p&gt;
&lt;p&gt;I was a little bit shocked that during the 4.2 merge window of Linux, TCP too got a new congestion control algorithm that&amp;rsquo;s a little bit similar to the one I used in my protocol - that is, based on the change rate of RTT, rather than based on a packet lost.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
