in> -----Original Message----- > From: Daniel Quinlan [mailto:[EMAIL PROTECTED] > Sent: Thursday, July 08, 2004 10:46 PM > To: Johannes russek > Cc: Michael Parker; [EMAIL PROTECTED] > Subject: Re: AW: libspamc > > > "Johannes russek" <[EMAIL PROTECTED]> writes: > > > hi michael. maybe you're right about a pure perl implementation, but > > for me it seems that writing a perl extension to libspamc is way less > > time consuming, for example all that ssl stuff needs to be > > reprogrammed. anyway, most systems should have a c compiler, and > > looking over CPAN, this assumption is made very often :) are you > > interested in my work anyway, when i'm finished? (will be this days) > > It's more common for non-Unix installs (read: Windows) to have Perl, but > no installed C compiler. It's okay for a pure Perl implementation to > require modules like IO::Socket::SSL, of course.
ah, okay. the windows guys. a pure perl interface would of course be very platform independent, but i wonder if the performance impact would be big or not, as performance might be an important issue for the ones using spamc/spamd. i have got a pure perl implementation of the spamc part, but it does only support UNIX domain sockets, not tcp. that might not be very hard to implement, but we could try to bench my xs interface against that pure perl thingie on a socketfile, when i'm finished. > > So, yes, we're interested in the work anyway, but I'd drool over a pure > Perl implementation. Not only would it add something missing on some > systems, but it would also make it easy to test and experiment with > changes and it could be used as a cross-check of the C implementation in > our "make test". > > Daniel > > -- > Daniel Quinlan > http://www.pathname.com/~quinlan/ > anyway, another question to the list about this issue: is it planned to have libspamc include more details about the tests spamd made? right now there is only the report and the score/threshold, nothing else. so if one wants to have the number of hits, the names of the tests and so on, he (well, in this case, me :)) has to regexp it out of the report. that could have another performance impact, so why don't make the spamd protocoll more machine processable and add HTTP like some lines like TestsHit: RAZOR2_CHECK DCC_CHECK DNS_FROM_RFC1_DNS and so on. could that be done? regards, johannes
