I am repeatedly encountering connection problems. so what you recommend me 
to fix it? 

On Saturday, May 9, 2020 at 12:02:25 AM UTC+1 [email protected] wrote:

>
>
> On Wednesday, February 3, 2016 at 4:42:59 PM UTC-5, David Benjamin wrote:
>>
>> *Primary eng (and PM) emails*
>> [email protected], [email protected]
>>
>> *Summary*
>> TLS has a version negotiation mechanism to securely introduce new 
>> versions without breaking compatibility. Yet some buggy servers (usually 
>> called “version-intolerant”) implemented this wrong, so browsers were 
>> forced to add insecure fallbacks to work around this. On failed 
>> connections, we retry, advertising lower and lower maximum versions.
>>
>> The SSL 3.0 fallback leg was removed last year due to POODLE (followed by 
>> removing SSL 3.0 entirely). Then in Chrome 45, we removed another leg 
>> <https://groups.google.com/a/chromium.org/d/msg/security-dev/F6ZjP6FnyRE/bK7TKtvnHYsJ>
>>  as 
>> a trial run to removing the whole thing. We would like to finish the job 
>> and remove it completely.
>>
>> *This does not remove support for TLS 1.0 or TLS 1.1.* Although servers 
>> are urged to upgrade to TLS 1.2 (all ciphers in prior versions have known 
>> problems), Chrome will continue to support TLS 1.0 and TLS 1.1 at this 
>> time. Only buggy servers which do not correctly implement TLS will be 
>> affected.
>>
>> *Motivation*
>> The fallback bypasses TLS's downgrade protection. Attackers can force *all 
>> sites* (not just the buggy ones) to negotiate weaker versions, despite 
>> both client and server supporting newer, more secure protocol versions. 
>> This causes problems when attacks are found against old protocol versions, 
>> such as POODLE. FALLBACK_SCSV (RFC 7507 
>> <https://tools.ietf.org/html/rfc7507>) fixes this, but it requires both 
>> client and server support.
>>
>> TLS fallbacks also mask implementation bugs. We’ve seen many examples of 
>> vendors shipping new server software with bugs that could have been noticed 
>> were it not for the fallback.
>>
>> *Compatibility risk*
>> Any remaining 1.2-intolerant sites will break. After Chrome 45 removed 
>> all but the TLS 1.1 fallback, these must be 1.2-intolerant but not 
>> 1.1-intolerant. We are not aware of any such servers. (One would hope these 
>> bugs have been resolved as this is the third version bump since SSL 
>> 3.0.) Mozilla has already removed the fallback as of Firefox 37 
>> <https://www.mozilla.org/en-US/firefox/37.0/releasenotes/>, so any 
>> 1.2-intolerant sites would not load in Firefox.
>>
>> There is an additional risk that some unrelated server bug was masked by 
>> the fallback and doesn’t manifest in other browsers. Notably, we are aware 
>> of an issue <https://crbug.com/433406> affecting Chrome connections to a 
>> small fraction of Windows-based servers, though it appears much less 
>> prevalent than previously believed. (Metrics below.)
>>
>> *Alternative implementation suggestion for administrators*
>> Affected sites will fail to load with a dedicated network error code, 
>> ERR_SSL_FALLBACK_BEYOND_MINIMUM_VERSION. (This page already exists from 
>> previous iterations of this.) We’ll link the error page to a Help Center 
>> article with appropriate instructions for administrators:
>>
>> Server administrators should ensure their software is up-to-date and, if 
>> still unresolved, reach out to their software vendor. Affected software 
>> vendors should fix bugs in their server implementations and follow the 
>> protocol specification. The Windows issue above was addressed in 
>> KB3042058 <https://support.microsoft.com/en-us/kb/3042058>. Note that 
>> KB3042058 describes important prerequisites that must be installed prior to 
>> installing KB3042058.
>>
>> For enterprise administrators, where our metrics are underrepresented, 
>> the SSLVersionFallbackMin administrative policy will be temporarily 
>> available to give more time to fix servers.
>>
>> *Usage information from UMA*
>> Net.ConnectionUsedSSLVersionFallback2 measures HTTPS requests which 
>> required the fallback. Over the past week, 0.0017% of HTTPS requests 
>> required the fallback.
>>
>> A few releases ago, in Chrome 46, the number was 0.015%. The difference 
>> was previously believed to be the Windows bug, but it seems to have been a 
>> different issue which is no longer a concern. (We experimentally worked 
>> around the bug, and numbers went down. Then we deprecated DHE 
>> <https://groups.google.com/a/chromium.org/d/msg/security-dev/dYyhKHPnrI0/pNxx8vTKBAAJ>
>>  for 
>> unrelated reasons, and then removed the experiment, but numbers stayed 
>> down. We've since observed a DHE-specific issue, also affected by the 
>> workaround, which we believe was the source.)
>>
>> *OWP launch tracking bug*
>> https://crbug.com/583787
>>
>> *Entry on the feature dashboard*
>> https://www.chromestatus.com/feature/5685183936200704
>>
>> *Requesting approval to remove too?*
>> Yes.
>>
>> As an extra precaution, the removal will be gated on field trial, so if 
>> anything unexpected occurs, we can easily roll back. We’ll also continue to 
>> monitor metrics leading up to Chrome 50 reaching stable.
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/a/chromium.org/d/msgid/blink-dev/e5abad05-f1c9-473b-baca-c05a23c9a008n%40chromium.org.

Reply via email to