Re: SSL Bump differences at various steps
Amos Jeffries <[email protected]> Tue, 10 Mar 2026 21:41:10 +1300
| Newsgroups | gmane.comp.web.squid.general |
|---|---|
| Message-ID | <[email protected]> |
On 06/03/2026 22:35, Andrey K wrote: > Hello, > > Iwaswonderingwhatare the differencesif we apply the sslbump > operationatdifferentsteps? > The documentation contains information about only one difference - when > we bump at the step1 Proxy first establishes a TLS connection with the > Client, and then with the Server, while in the other cases - first with > the Server, and then with the Client. > > I looked into this issue and tried to summarize the information I have > (I will not detail here the information available to the admin at each > of the three steps). Maybe it will help someone. > Thank you. Though, for anyone reading this in the future - be aware that Squid behaviour which is _not_ documented officially is subject to change without notice. There are still bugs being found and removed from this feature. For example, see below... > Bumping at steps 2 and 3 are very similar,butinthe firstcase, > ProxysendsClientHelloto a Server withitsownciphers,andinthe lattercase - > withciphers receivedfromthe Client. > > If anyone has any comments or additions, please feel free to complete it. > This looks like a bug to me. Squid should only be preserving the client ciphers etc when "peek" is used - in order to permit step2/3 splice. The explicit configuration of "stare" in your tests should be enabling Squid to filter the ciphers it sends to make your test #2 and #3 identical traffic. What your test #2 is showing is what I would expect from the slightly weird configuration: ssl_bump peek step1 ssl_bump stare step2 ssl_bump bump step3 or just, ssl_bump stare step2 ssl_bump bump step3 Cheers Amos