On Dec 27, 2010, at 5:23 AM, Daniel Stenberg wrote:

> I think it will make sense to still get some #define set internally in 
> libcurl when TLSAUTH support is found and used, so that we can make 
> curl_easy_setopt() return failure for the cases where an app would try to set 
> the TLSAUTH options without there being underlying support for them.

Got it. Will work on this.

>> I'll also work on some tests (will have to figure out how to get stunnel 
>> working with TLS-SRP).
> 
> Cool! But since stunnel is OpenSSL-based, won't this require that we build 
> stunnel with an OpenSSL with the SRP patch applied?

Yes, didn't think about this. What do you recommend the tests work against? The 
only server implementations of TLS-SRP that don't require patching are 
Apache/mod_gnutls and TLS Lite (a Python lib), unless I'm missing any. 
Apache/mod_gnutls is too heavy to use for testing. Given the lack of better 
options, is it OK to add a dependency of Python and TLS Lite for the TLS-SRP 
tests?

(I've talked to some people about the OpenSSL TLS-SRP patch. It has been 
maintained and works against recent OpenSSL releases, but it's still unclear 
when it will be accepted.)

-Quinn
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html

Reply via email to