On Wed, 4 Feb 2026 19:29:02 GMT, Kirill Shirokov <[email protected]> wrote:

> Removed FFDHE_6144 and FFHDE_8192 from the default list of TLS named groups, 
> so now to consider them as candidates in TLS handshake user has to enable 
> them explicitly (e.g. `-Djdk.tls.namedGroups=ffdhe6144,ffhde8192`)
> 
> Tested on Linux x64/aarch64, MacOS aarch64, Windows x64 using jtreg 
> `test/jdk/sun/security/ssl` and `test/jdk/javax/net/ssl`.
> 
> [tests-linux-aarch64.log](https://github.com/user-attachments/files/25080233/tests-linux-aarch64.log)
> [tests-linux-x86.log](https://github.com/user-attachments/files/25080235/tests-linux-x86.log)
> [tests-macos-aarch64.log](https://github.com/user-attachments/files/25080236/tests-macos-aarch64.log)
> [tests-windows-x64.log](https://github.com/user-attachments/files/25080237/tests-windows-x64.log)

I agree with Sean.  I can't think of a case where these larger DHE exchanges 
are used.  When we did hybrid named groups, we looked at what many browsers and 
libraries did.  While browsers like Mozilla and libraries like OpenSSL still 
have FFDHE named groups, they all use `ffdhe2048` and `ffdhe3072`, nothing 
larger.  We'd leave `ffdhe4096` for the more stringent security cases, but 
anything beyond that is just not seen out there AFAICT.  And we're just talking 
about the default case here - in some particularly security-sensitive case 
users can still enable these via property or API calls (putting aside the 
question of why something that security sensitive would want DHE at all when 
hybrid named groups are available and larger EC solutions like secp521r1 and 
x448).

-------------

PR Comment: https://git.openjdk.org/jdk/pull/29577#issuecomment-3855577496

Reply via email to