<div dir="ltr"><br>On Tue, Dec 12, 2017 at 9:47 PM Dave Taht <<a href="mailto:dave.taht@gmail.com">dave.taht@gmail.com</a>> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
In a world where quic is 7% of all traffic presently (wow, I didn't know)...<br>
<br></blockquote><div><br></div>BitTorrent's MicroTP (uTP) had about 15% at some point as well. This was especially scary as:</div><div class="gmail_quote"><br></div><div class="gmail_quote">* it used the then not-fully-understood LEDBAT congestion algorithm.</div><div class="gmail_quote">* the de-facto implementation wasn't too well understood and there no competing implementations in order to understand interoperability.</div><div class="gmail_quote">* there were no formal process when it was created.</div><div class="gmail_quote">* for simplicty it dropped half-open connections. But implemented that in a slightly faulty way.</div><div class="gmail_quote"><br></div><div class="gmail_quote">I know, because I had an experiment where I did an Erlang-implementation of the protocol.</div><div class="gmail_quote"><br></div><div class="gmail_quote">At least QUIC seems to be getting some process now, which is good.<br></div></div>