I remember a big discussion here about binary prompts with pam, and how that may solve problem with integrating pam to an applications uses some protocol (like pop or ftp) that limits pam usage. I still can't understand what are binary prompts and how them may solve a problem (can someone enlighten me here?). But I used pop daemon (written by me) here that implements APOP command using pam. When APOP is enabled and daemon gets it from client, it then sets up pam_env with it's random seed (APOP_SEED=012345), sets up PAM_USER as usual, and uses apop hash value as PAM_AUTHTOK when requested. In pop's pam stack, I defined "pam_apop" module as sufficient (first in a stack) that checks APOP_SEED variable and verifies supplied hash if variable is set, returning OK or AUTHFAIL, or returns PAM_IGNORE if not set. This isn't perfect (pam_apop must be the first module who asks for an AUTHTOK, for example), but it works. The question: how this may be done with binary prompts (if can at all -- note that at least generation of random seed should be done at pop server startup -- it should be shown in a first welcome line server responds to connection attempt)? Note that APOP is only example here -- think of any other AUTH in pop/imap, or even STLS/STARTTLS - should that last one fit with pam also (STARTTLS may provide user's certificate info)? The question is real: if e.g. APOP can be done with binary prompts, then there is no need to introduce other "protocols" around pam like APOP_SEED environment variable, but I'm afraid that if I release my pop daemon to public as it is now, people will start to write pop modules for it based on wrong "pam protocol". Thank you! Regards, Michael.
