current/doc/source faq.sgml,2.117,2.118
Fabian Keil <[email protected]>
| Newsgroups | gmane.comp.web.privoxy.cvs |
|---|---|
| Message-ID | <[email protected]> |
Update of /cvsroot/ijbswa/current/doc/source In directory sfp-cvs-1.v30.ch3.sourceforge.com:/tmp/cvs-serv25508/doc/source Modified Files: faq.sgml Log Message: Add a FAQ entry for tainted sockets Index: faq.sgml =================================================================== RCS file: /cvsroot/ijbswa/current/doc/source/faq.sgml,v retrieving revision 2.117 retrieving revision 2.118 diff -C2 -d -r2.117 -r2.118 *** faq.sgml 4 Mar 2016 13:21:33 -0000 2.117 --- faq.sgml 6 Mar 2016 11:32:06 -0000 2.118 *************** *** 2953,2956 **** --- 2953,2992 ---- </sect2> + <sect2 renderas="sect3" id="tainted-sockets"> + <title>What are tainted sockets and how do I prevent them?</title> + <para> + &my-app; marks sockets as tainted when it can't use them to + serve additional requests. + This does not necessarily mean that something went wrong and + information about tainted sockets is only logged if connection + debugging is enabled (debug 2). + </para> + <para> + For example server sockets that were used for CONNECT requests + (which are used to tunnel https:// requests) are considered tainted + once the client closed its connection to &my-app;. + Technically &my-app; could keep the connection to the server open, + but the server would not accept requests that do not belong to the + previous TLS/SSL session (and the client may even have terminated + the session). + </para> + <para> + Server sockets are also marked tainted when a client requests a + resource, but closes the connection before &my-app; has completely + received (and forwarded) the resource to the client. + In this case the server would (probably) accept additional requests, + but &my-app; could not get the response without completely reading + the leftovers from the previous response. + </para> + <para> + These are just two examples, there are currently a bit more than + 25 scenarios in which a socket is considered tainted. + </para> + <para> + While sockets can also be marked tainted as a result of a technical + problem that may be worth fixing, the problem will be explicitly + logged as error. + </para> + </sect2> </sect1> ------------------------------------------------------------------------------