[master] 79d1c54 A new, updated SSL rant.

[Poul-Henning Kamp]

commit 79d1c54ccafd8920e208ea3c433281cbb120c843
Author: Poul-Henning Kamp <phk at FreeBSD.org>
Date:   Tue Apr 28 11:34:34 2015 +0000

    A new, updated SSL rant.

diff --git a/doc/sphinx/phk/index.rst b/doc/sphinx/phk/index.rst
index 3cef4af..425d4e5 100644
--- a/doc/sphinx/phk/index.rst
+++ b/doc/sphinx/phk/index.rst
@@ -8,6 +8,7 @@ You may or may not want to know what Poul-Henning thinks.
 .. toctree::
 	:maxdepth: 1
 
+	ssl_again.rst
 	persistent.rst
 	dough.rst
 	wanton_destruction.rst
diff --git a/doc/sphinx/phk/ssl_again.rst b/doc/sphinx/phk/ssl_again.rst
new file mode 100644
index 0000000..ba4ecf1
--- /dev/null
+++ b/doc/sphinx/phk/ssl_again.rst
@@ -0,0 +1,154 @@
+.. _phk_ssl_again:
+
+=============
+SSL revisited
+=============
+
+Four years ago, I wrote a rant about why Varnish has no SSL support
+(:ref:`phk_ssl`) and the upcoming 4.1 release is good excuse to
+revisit that issue.
+
+A SSL/TLS library
+~~~~~~~~~~~~~~~~~
+
+In 2011 I critized OpenSSL's source-code as being a nightmare,
+and as much as I Hate To Say I Told You So, I Told You So:  See also
+"HeartBleed".
+
+The good news is that HeartBleed made people realize that FOSS
+maintainers also have mortgages and hungry kids.
+
+Various initiatives have been launched to make prevent critical
+infrastructure software from being maintained sunday evening between
+11 and 12PM by a sleep-deprived and overworked parent, worried about
+about being able to pay the bills come the next month.
+
+We're not there yet, but it's certainly getting better.
+
+However, implementing TLS and SSL is stil insanely complex, and
+thanks to Edward Snowdens whistle-blowing, we have very good reasons
+to belive that didn't happen by accident.
+
+The issue of finding a good TLS/SSL implementation is still the
+same and I still don't see one I would want my name associated with.
+
+OpenBSD's LibreSSL is certainly a step in a right direction, but
+time will show if it is viable in the long run -- they do have
+a tendency to be -- "SQUIRREL!!" -- distracted.
+
+Handling Certificates
+~~~~~~~~~~~~~~~~~~~~~
+
+I still don't see a way to do that.  The Varnish worker-process is not
+built to compartementalize bits at a cryptographic level and making it
+do that would be a non-trivial undertaking.
+
+But there is new loop-hole here.
+One night, waiting for my flight home in Oslo airport, I went though
+the entire TLS/SSL handshake process to see if there were anything
+one could do, and I realized that you can actually terminate TLS/SSL
+without holding the certificate, provided you can ask some process
+which does to do a tiny bit of work.
+
+The next morning `CloudFlare announced the very same thing`_:
+
+.. _CloudFlare announced the very same thing: https://blog.cloudflare.com/keyless-ssl-the-nitty-gritty-technical-details/
+
+This could conceiveably be a way to terminate TLS/SSL in the Varnish-worker
+process, while keeping the most valuable crypto-bits away from it.
+
+But it's still a bad idea
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+As I write this, the news that `apps with 350 million downloads`_ in total
+are (still) vulnerable to some SSL/TLS Man-In-The-Middle attack is doing the
+rounds.
+
+.. _apps with 350 million downloads: http://arstechnica.com/security/2015/04/27/android-apps-still-suffer-game-over-https-defects-7-months-later/
+
+Code is hard, crypto code is double-plus-hard, if not double-squared-hard,
+and the world really don't need another piece of code that does an
+half-assed job at cryptography.
+
+If I, or somebody else, were to implement SSL/TLS in Varnish, it would
+talk at least half a year to bring the code to a point where I would be
+willing to show it to the world.
+
+Until I get my time-machine working, that half year would be taken
+away of other Varnish development, so the result had better be worth
+it: If it isn't, we have just increased the total attack-surface
+and bug-probability for no better reason than "me too!".
+
+When I look at something like Willy Tarreau's `HAProxy`_ I have a
+hard time to see any significant opportunity for improvement.
+
+.. _HAProxy: http://www.haproxy.org/
+
+
+Conclusion
+~~~~~~~~~~
+
+No, Varnish still won't add SSL/TLS support.
+
+Instead in Varnish 4.1 we have added support for Willys `PROXY`_
+protocol which makes it possible to communicate the extra details
+from a SSL-terminating proxy, such as `HAProxy`_, to Varnish.
+
+.. _PROXY: http://www.haproxy.org/download/1.5/doc/proxy-protocol.txt
+
+From a security point of view, this is also much better solution
+than having SSL/TLS integrated in Varnish.
+
+When (not if!) the SSL/TLS proxy you picked is compromised by a
+possibly planted software bug, you can pick another one to replace
+it, without loosing all the benefits of Varnish.
+
+That idea is called the "Software Tools Principle", it's a very old
+idea, but it is still one of the best we have.
+
+
+Political PostScript
+~~~~~~~~~~~~~~~~~~~~
+
+I realize that the above is a pretty strange stance to take in the
+current "SSL Everywhere" political climate.
+
+I'm not too thrilled about the "SSL Everywhere" idea, for a large
+number of reasons.
+
+The most obvious example is that you don't want to bog down your
+countrys civil defence agency with SSL/TLS protocol negotiations,
+if their website is being deluged by people trying to survive a
+natural disaster.
+
+The next big issue is that there are people who do not have a right
+to privacy.  In many countries this includes children, prisoners,
+stock-traders, flight-controllers, first responders and so on.
+
+SSL Everywhere will force institutions to either block any internet
+connectivity or impose Man-in-The-Middle proxies to comply with
+legal requirements of logging and inspection.  A clear step in
+the wrong direction in my view.
+
+But one of the biggest problem I have with SSL Everywhere is that
+it gives privacy to the actors I think deserve it the least.
+
+Again and again shady behaviour of big transnational, and therefore
+law-less, companies have been exposed by security researchers (or
+just interested lay-people) who ran tcpdump. snort or similar traffic
+capture programs and saw what went on.
+
+Remember all the different kind of "magic cookies" used to track
+users across the web, against their wish and against laws and regulations ?
+
+Pretty much all of those were exposed with trivial packet traces.
+
+With SSL Everywhere, these actors get much more privacy to invade
+the privacy of every human being with an internet connection, because
+it takes a lot more skill to look into a SSL connection than a
+plaintext HTTP connection.
+
+"Sunshine is said to be the best of disinfectantants" wrote supreme
+court justice Brandeis, SSL Everywhere puts all traffic in the shade.
+
+Poul-Henning, 2015-04-28