InCommon Baseline TLS 1.2

[Peter Schober]

* IAM David Bantz <dabantz at alaska.edu> [2021-06-30 23:32]:
> TLS 1.0 (1999) and TLS 1.1 (2006) are formally deprecated by IETF RFC 8996.
> 
> These versions lack support for current and recommended cryptographic
> algorithms and mechanisms, and various government and industry
> profiles of applications using TLS now mandate avoiding these old TLS
> versions. TLS version 1.2 became the recommended version for IETF
> protocols in 2008...

I don't think the question was whether it was "secure" to keep
offering TLS 1.0 and/or TLS 1.1. ;)

> Web sites that negotiate a TLS 1.0 or 1.1 protocol will trigger user
> warnings that connections are “not secure” from Chrome and other browsers.

>From modern releases of those browsers, maybe. And I'm not convinved
even that's true: You're saying a modern Chrome or FF release will
warn even if the server in question also speaks an accaptable TLS
version? (I've certainly never encountered that myself.)

But modern browser releases speak TLS 1.2 and TLS 1.3 and so are not
really the main concern when considering the removal of older TLS
versions. (Unless the browser vendors are in fact using new releases
to "shame" servers also offering older versions, as per above.)

> There are some niche needs for the older protocols to support legacy
> devices that cannot support newer secure TLS, but you can support
> legacy clients such as IE 11 and Android 5 using TLS 1.2.

FWIW and not Shib-related, on one of our more widely (literally,
including people from outside our own or even the academic community)
used services I had first enabled TLS version logging and later
disabled TLS 1.0 (so I don't see those requests anymore as they are
being rejected by the web server).  Checking for TLS 1.1 request I
only see a single legit one in recent weeks, and yes (@Scott), coming
from one of our medical universities ("classic" Firefox 52, in this
case, if you believe the U-A string). ;)

The comments from Scott about end user support and management are of
course spot on (as always) but also the hardest to change if your SOL.
  Unless you're under severe time pressure to act monitoring what TLS
versions your server is actually seeing for a couple of weeks should
give you a better picture of the consequences, though of course
susceptible to the Problem of Induction[1].

-peter

[1] https://en.wikipedia.org/wiki/Problem_of_induction