Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Jacob Champion
On 08/25/2016 03:34 PM, William A Rowe Jr wrote: My 2c... Exclusion lists are far preferable to allow lists. .conf files seem to persist for a decade and longer. There is no anticipating what will be added to the list of unwise ciphers a year from now, but that goes for an explicit list or for a

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Yann Ylavic
On Fri, Aug 26, 2016 at 12:56 AM, Yann Ylavic wrote: > On Fri, Aug 26, 2016 at 12:34 AM, William A Rowe Jr > wrote: >> >> Exclusion lists are far preferable to allow lists. .conf files seem to >> persist for >> a decade and longer. There is no anticipating what will be added to the list >> of un

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Yann Ylavic
On Fri, Aug 26, 2016 at 12:34 AM, William A Rowe Jr wrote: > > Exclusion lists are far preferable to allow lists. .conf files seem to > persist for > a decade and longer. There is no anticipating what will be added to the list > of unwise ciphers a year from now, but that goes for an explicit list

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Jacob Champion
On 08/25/2016 03:37 PM, Yann Ylavic wrote: Note that this thread recommends: DEFAULT:!EXPORT:!LOW:!MEDIUM which, with openssl 1.1, selects DHE-RSA-AES256-SHA before e.g. ECDHE-ECDSA-CHACHA20-POLY1305. So some tuning is needed there too... Yeah, I honestly don't agree with Viktor's blanket rec

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Yann Ylavic
On Thu, Aug 25, 2016 at 11:37 PM, Yann Ylavic wrote: >>> >>> Actually, intermediate looks more like: >>> kECDHE:kDHE:kRSA:+SHA:!MEDIUM:!LOW:!aNULL:!eNULL:!DSS:!RC4:!3DES > > The CipherSuite above is perfectly fine with all versions AFAICT... I spoke too quickly, libressl does not understand the k

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread William A Rowe Jr
On Thu, Aug 25, 2016 at 5:09 PM, Jacob Champion wrote: > On 08/25/2016 02:37 PM, Yann Ylavic wrote: > >> I find this CipherSuite quite evolutive and unsurprising (key exchange >> algorithms don't change or are introduced too often, that's an >> euphemism :), if a cipher proves to be weak, add it

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Jacob Champion
On 08/25/2016 02:37 PM, Yann Ylavic wrote: I find this CipherSuite quite evolutive and unsurprising (key exchange algorithms don't change or are introduced too often, that's an euphemism :), if a cipher proves to be weak, add it to the :!END and be done (like RC4 and 3DES recently). To be clear

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Yann Ylavic
On Thu, Aug 25, 2016 at 11:04 PM, Jacob Champion wrote: > On 08/25/2016 01:44 PM, Yann Ylavic wrote: >> >> On Thu, Aug 25, 2016 at 10:26 PM, Yann Ylavic >> wrote: >>> >>> An exhaustive ciphers list looks not evolutive to me, and depends on >>> the SSL library version. >>> >>> "Modern" ciphers cou

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Jacob Champion
On 08/25/2016 02:04 PM, Jacob Champion wrote: (HIGH was supposed to be the evolutive way to go, but IIRC that failed due to backwards compatibility concerns when OpenSSL tried to remove the weak ciphers from it.) (For more exciting reading on the cipher compatibility saga, see https://mta.o

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Jacob Champion
On 08/25/2016 01:44 PM, Yann Ylavic wrote: On Thu, Aug 25, 2016 at 10:26 PM, Yann Ylavic wrote: An exhaustive ciphers list looks not evolutive to me, and depends on the SSL library version. "Modern" ciphers could possibly be defined by 'kECDHE:!MEDIUM:!LOW:!aNULL:!eNULL:!SSLv3', and "Intermedi

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Jacob Champion
Thanks for the review! On 08/25/2016 12:50 PM, Eric Covener wrote: - If we talk about BREACH we can't just show "SSLCompression off" because BREACH, IIUC, would affect deflate over TLS not just TLS compression. Right, `SSLCompression off` is there to address the general CRIME-type vulnerabil

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Yann Ylavic
On Thu, Aug 25, 2016 at 10:26 PM, Yann Ylavic wrote: > Hi Jacob, > > On Thu, Aug 25, 2016 at 7:36 PM, Jacob Champion wrote: >> >> If you're interested, the relevant commit is r1757280 in httpd. > > An exhaustive ciphers list looks not evolutive to me, and depends on > the SSL library version. > >

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Yann Ylavic
Hi Jacob, On Thu, Aug 25, 2016 at 7:36 PM, Jacob Champion wrote: > > If you're interested, the relevant commit is r1757280 in httpd. An exhaustive ciphers list looks not evolutive to me, and depends on the SSL library version. "Modern" ciphers could possibly be defined by 'kECDHE:!MEDIUM:!LOW:!

Re: Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Eric Covener
On Thu, Aug 25, 2016 at 1:36 PM, Jacob Champion wrote: > If you're interested, the relevant commit is r1757280 in httpd. It looks reasonable / conventional wisdom to me. Only things that stuck out to me: - If we talk about BREACH we can't just show "SSLCompression off" because BREACH, IIUC, wo

Requesting review for SSL how-to changes (r1757280)

2016-08-25 Thread Jacob Champion
Hi all, I've been informed that docs backports are CTR, but since this is my first time updating them (and it's a security document that I've updated), I was hoping for a couple more eyes. If you're interested, the relevant commit is r1757280 in httpd. Thanks! --Jacob --