+# We get (TLS1.3 , OpenSSL):
+# SYN >
+# < SYN,ACK
+# ACK >
+# Client Hello >
+# < Server Hello, Change Ciph, Extensions, Cert, Cert Verify, Finished
+# Change Ciph,Finsh >
+# < Banner
+# EHLO >
+# < EHLO resp
+# MAIL,RCPT,DATA >
+# < ACK,ACK,DATA-go-ahead
+#
+# GnuTLS splits both the server records and the client response pair over two TCP segments:
+# Client Hello >
+# < Server Hello, Change Ciph
+# Change Ciph >
+# < Extensins, Cert, Cert Verify, Finished
+# Finished >
+# (otherwise the same). The extra segments are piplined and do not incur an extra roundtrip time.
+#
+# To see that pipelining:
+# sudo tc qdisc add dev lo root netem delay 50ms / sudo tc qdisc delete dev lo root
+#
+# To test TFO, enable in the transport in the conf/ file
+# With TFO we get the Client Hello on the SYN, and the initial Server segment pipelined with/after the SYN,ACK
+# and before the 3rd-ACK. We still can't merge the 3rd-ACK with the second Client record set,
+# but it does ack the initial Server data.
+#
+# To see the TFO((R):
+# First clear any previously-obtained cookie:
+#sudo perl
+#open(INFO, "-|", "/usr/bin/uname -s");
+#$_ = <INFO>;
+#if (/^FreeBSD/) {
+#system("sysctl net.inet.tcp.fastopen.client_enable=0"); system("sysctl net.inet.tcp.fastopen.client_enable=1");
+#} else {
+#system ("[ -e /proc/sys/net/ipv4/tcp_fastopen_blackhole_timeout_sec ] && echo 0 > /proc/sys/net/ipv4/tcp_fastopen_blackhole_timeout_sec");
+#system ("ip tcp_metrics delete 127.0.0.1");
+#}
+#
+#****
+#
+#