Hi Phil, >> 2. What's the rationale for moving the header? I prefer writing #include >> <zmq.h> to #include<zmq/zmq.h> personally. > > The primary rational is to provide a dedicated location in which developers > can inspect the headers and view the api. One of the first things I do when > working with libraries is turn to the include area and view those headers. > > One cannot read the headers if one does not know what they are. Providing > a private directory for them is a convenience to developers. > > When I look at my include area, I see: > > $ ls -1 z* > > zclock.h > zconf.h > zctx.h > zfile.h > zframe.h > zhash.h > zlib.h > zlist.h > zloop.h > zmq.h > zmq.hpp > zmq_utils.h > zmsg.h > zsocket.h > zsockopt.h > zstr.h > zthread.h > > It takes a keen eye to discern that zlib.h is not part of your libraries. > Worse, > suppose one day you wish to add a header called 'zlib.h'? Having to worry > about > name collisions across the universe of headers is unreasonable with so many > packages > out there these days.
Ok, right. It's messy. However, note that the only header from the list that belongs to core 0MQ project is zmq.h. The rest are from czmq project so I guess the fix should be done there. What you would get is a single header file in /usr/local/include (zmq.h) and a directory czmq that would contain the remaining czmq-related headers. Placing a single zmq.h file into a subdirectory seems to be an over-kill IMO. Thoughts? Martin _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
