Hello, I strongly urge the OpenSC team to make this change to fix mini-driver compatibility. Yes, my company "needs this because we just do", however by implementing this fix, it will ensure better compatibility with a Windows mini-driver and thus a more usable product for more people.
Thanks, Brian Thomas -----Original Message----- From: Martin Paljak [mailto:mar...@martinpaljak.net] Sent: Saturday, January 22, 2011 7:27 AM To: Douglas E. Engert Cc: opensc-devel@lists.opensc-project.org; Brian Thomas Subject: Re: [opensc-devel] Proposed cardmod patch Hello, On Jan 21, 2011, at 9:45 PM, Douglas E. Engert wrote: > Attached is the patch to just do (2) below. I believe that this > is the part that Brian wants to not call sc_ctx_detect_readers > Martin you had said you want to do this differently so I will > leave this up to you as to weither this is commited or not. > On 1/19/2011 5:46 PM, Douglas E. Engert wrote: >> (2) Implement a change that this should work for Brian Thomas >> that allows one to not detect the reader when creating >> the sc_context. This uses the >> SC_CONTEXT_PARAM_DONT_DETECT_READERS flag in the ctx_params >> >> (The more I think about this, this should be the default >> to not detect the readers during the sc_create_ctx, as >> a flag on opensc.conf could control this instead.) Before adding any flags, the real reason for it should be identified. If the reason is behavior of OpenSC internals when used for a MS MiniDriver implementation, the requirements should be clear and communicatable: * context and card handles are handed in by the minidriver code instead of OpenSC having to detect them itself during context creation (through context creation parameters of course) * detecting of readers other than the given reader make no sense and/or can cause unwanted side effects (side effects and bugs should be fixed separately or not allowed function chaining be documented) * what happens or should not happen during sc_context_create() has not been documented anywhere (but should, no matter which of the following is implemented) I suggest to a) remove the flags member from context parameters (until there is a need for flags) b1) remove the call to sc_detect_readers() in sc_context_create() and document the behavior of sc_context_create(), delegating reader detection responsibility to the application (command line utility or API bridge like PKCS#11 or Tokend or CSP) OR b2) based on the passed in parameters, behave as expected when context and card handle are passed in (don't list readers, fetch the reader properties (name, protocol) from the card handle etc), extending reader-pcsc.c if needed to support a different initialization method. Normal context creation would still behave the same way as currently, by detecting available readers during context creation time. (eventually c)) fix reader-pcsc.c by removing the cardmod copy case and adjusting reader-pcsc.c to to handle the restrictions of minidriver use (code restrictions based on the mode of driver initialization which can further be narrowed to Windows only ifdefs) I don't really have a preference, either is good, with b1 being a simpler change at the moment but b2 combined with c being necessary in the long run. If the need is because of something different and the above does not fit, Brian needs to come up with a better excuse than "we need this because we just do" ;) Cheers, -- @MartinPaljak.net +3725156495 _______________________________________________ opensc-devel mailing list opensc-devel@lists.opensc-project.org http://www.opensc-project.org/mailman/listinfo/opensc-devel
