> > > > * 47: There should only be a "libfoo.h" when there is a > > corresponding > > > > libfoo.so. This header file should be named something more > > > > appropriate like "ofmt.h". Likewise, fix comment on line 77 > > > > > > I was following the precedent set by libdladm/common/lib* (many files > > > of which were added by Clearview :-) > > > > The libdladm header collection was a bit of a compromise and in retrospect > > a confusing one -- but at least everything beings with "libdl". > > True. > > FWIW, my personal opinion is that the collection of header files that > libdladm API consumers need to pick from is a mistake (not just a > confusing compromise), and one that's not too late to correct. > Consumers of the libdladm API (regardless of what corner of the API > they're using) should only need to include a single header file. All > consumers need to include libdladm.h in order to call dladm_open(), > which is a pre-requisite to using any other part of the API, having all > symbols visible from there is much simpler. It doesn't pollute the > application namespace either, since libdladm symbols are carefully > prefixed with "dladm_".
I think this is a judgment call based on the usage patterns and scale of the library. For instance, I don't think I'd want "libc.h" ;-) I agree this isn't too late to fix but I also don't find it a raging fire. -- meem
