ORLY

I seem to recall this happening on 10.10 as well, but it could be an El Capitan 
thing. Do you mind reminding me of your squid config Jason?

Thanks!

> On 16 Oct 2015, at 11:06 AM, Jason Haar <jason_h...@trimble.com> wrote:
> 
> Just a data point, but I've just got up Safari on Yosemite connecting
> through squid-3.5.10 to https://wikipedia.org/ with full bump-ing with
> no problems.
> 
> Same with twitter.com and github.com. Click on the padlock shows the
> server cert chaining to my squidCA cert (which is trusted of course)
> 
> ie this can't have anything to do with Elliptic Curves or pinning
> 
> Jason
> 
> On 15/10/15 12:19, Alex Rousskov wrote:
>> On 10/14/2015 05:00 PM, Dan Charlesworth wrote:
>> 
>>> I feel like if server-first is working there must be *some*
>>> combination of peek/stare/bump that’ll work too—it can’t be that
>>> “forward secrecy” cipher stuff.
>> 
>> While that feeling is natural, you should resist it. Newer SslBump
>> actions do not simply dissect the old ones into smaller steps. The old
>> actions (e.g., server-first) do not do some of the things that the new
>> actions do (e.g., peek extracts and sends SNI but server-first does
>> not). Doing more sometimes leads to more problems, especially in
>> experiment-driven features such as SslBump. Besides different cipher
>> negotiation patterns, you may be hitting a bug that server-first code
>> path lacks, for example.
>> 
>> 
>>> I really don’t want our customers to have to use server-first if they
>>> decide to employ bumping, so if any of you smart people have any
>>> other suggestions, please send them through.
>> 
>> I second Amos' implied suggestion to try the latest Squid 4.0 as the
>> next step. This does not mean you have to _deploy_ Squid 4.0:
>> 
>> * If Squid 4.0 does not work in your tests, we will not need to suspect
>> newer ciphers and may get more information from newer logs. We will also
>> be slightly more motivated to fix or improve something.
>> 
>> * If Squid 4.0 works, we will know more about your problem and may
>> suggest some other solutions if you have to run an older Squid.
>> 
>> In either case, do collect "debug_options ALL,9" cache logs for an
>> isolated test case.
>> 
>> 
>> Please note that I am not volunteering to examine your logs, and there
>> is no guarantee that this next step will lead to a solution, but it is
>> relatively easy to make that step.
>> 
>> 
>> HTH,
>> 
>> Alex.
>> 
>> 
>> 
>> 
>>>> On 15 Oct 2015, at 1:34 AM, Alex Rousskov 
>>>> <rouss...@measurement-factory.com> wrote:
>>>> 
>>>> On 10/13/2015 09:08 PM, Dan Charlesworth wrote:
>>>> 
>>>>> But in reality ssl_bump peek step1 & ssl_bump bump step3 is actually
>>>>> splicing everything, it seems.
>>>> 
>>>> This may not be related to your specific problem, but I want to clarify
>>>> the above.
>>>> 
>>>> ssl_bump peek step1
>>>> ssl_bump bump step3
>>>> 
>>>> A recent Squid mis-configured using the above sketch should indeed
>>>> splice everything. When Squid reaches bumping step2, no ssl_bump rule
>>>> matches, so Squid uses the previous step rule to decide what to do.
>>>> Since peeking implies splicing, Squid splices at step2 and never gets to
>>>> step3.
>>>> 
>>>> It is possible that, in his "bump at step3" recommendation below, Amos
>>>> was talking about this kind of configuration:
>>>> 
>>>> ssl_bump stare all
>>>> ssl_bump bump all
>>>> 
>>>> Bugs notwithstanding, the above results in bumping at step3.
>>>> 
>>>> Alex.
>>>> 
>>>> 
>>>>>> On 14 Oct 2015, at 1:51 PM, Amos Jeffries <squ...@treenet.co.nz> wrote:
>>>>>> 
>>>>>> On 14/10/2015 1:13 p.m., Dan Charlesworth wrote:
>>>>>>> Throwing this out to the list in case anyone else might be trying to 
>>>>>>> get SSL Bump to work with the latest version of Safari.
>>>>>>> 
>>>>>>> Every other browser on OS X (and iOS) is happy with bumping for pretty 
>>>>>>> much all HTTPS sites, so long as the proxy’s CA is trusted. 
>>>>>>> 
>>>>>>> However Safari throws generic “secure connection couldn’t be 
>>>>>>> established” errors for many popular HTTPS sites in including:
>>>>>>> - wikipedia.org
>>>>>>> - mail.google.com
>>>>>>> - twitter.com
>>>>>>> - github.com
>>>>>>> 
>>>>>>> But quite a number of others work, such as youtube.com.
>>>>>>> 
>>>>>>> This error gets logged to the system whenever it occurs:
>>>>>>> com.apple.WebKit.Networking: NSURLSession/NSURLConnection HTTP load 
>>>>>>> failed (kCFStreamErrorDomainSSL, -9802)
>>>>>>> 
>>>>>>> Apparently this is related to Apple’s new “App Transport Security” 
>>>>>>> protections, in particular, the fact that “the server doesn’t support 
>>>>>>> forward secrecy”. Even though it doesn’t seem to be affecting mobile 
>>>>>>> Safari on iOS 9 at all.
>>>>>>> 
>>>>>>> It’s also notable that Safari seems perfectly happy with legacy 
>>>>>>> server-first SSL bumping. 
>>>>>>> 
>>>>>>> I’m using Squid 3.5.10 and this is my current config: 
>>>>>>> https://gist.github.com/djch/9b883580c6ee84f31cd1
>>>>>>> 
>>>>>>> Anyone have any idea what I can try?
>>>>>> You can try bump at step3 (roughly equivalent to server-first) instead
>>>>>> of step2 (aka client-first).
>>>>>> 
>>>>>> 
>>>>>> Amos
>>>>>> 
>>>>>> _______________________________________________
>>>>>> squid-users mailing list
>>>>>> squid-users@lists.squid-cache.org
>>>>>> http://lists.squid-cache.org/listinfo/squid-users
>>>>> _______________________________________________
>>>>> squid-users mailing list
>>>>> squid-users@lists.squid-cache.org
>>>>> http://lists.squid-cache.org/listinfo/squid-users
>>> _______________________________________________
>>> squid-users mailing list
>>> squid-users@lists.squid-cache.org
>>> http://lists.squid-cache.org/listinfo/squid-users
>>> 
>> _______________________________________________
>> squid-users mailing list
>> squid-users@lists.squid-cache.org
>> http://lists.squid-cache.org/listinfo/squid-users
> 
> 
> -- 
> Cheers
> 
> Jason Haar
> Corporate Information Security Manager, Trimble Navigation Ltd.
> Phone: +1 408 481 8171
> PGP Fingerprint: 7A2E 0407 C9A6 CAF6 2B9F 8422 C063 5EBB FE1D 66D1
> 
> _______________________________________________
> squid-users mailing list
> squid-users@lists.squid-cache.org
> http://lists.squid-cache.org/listinfo/squid-users

_______________________________________________
squid-users mailing list
squid-users@lists.squid-cache.org
http://lists.squid-cache.org/listinfo/squid-users

Reply via email to