rja@edison.cho.ge.com (rja) (08/11/90)
Some articles here have recently raised the matter of what classes and library functions should be part of the ANSI C++ standard. I would hope that the ANSI working group would follow the lead of X3J11 as to what is appropriate for the language standard and what things are appropriate to leave to the IEEE 1003.* POSIX standards groups. The separation that IEEE 1003.1 & ANSI X3.159 achieved was really fairly good. Should there be a difference between "traditional" UNIX \(tm functions and semantics and those in approved POSIX standards that would actually affect C++ the language, I would hope that they would be resolved in favour of POSIX since all major vendours have committed to becoming POSIX compliant. I expect that there aren't any such problems since C++ is a language and hence should be defined in an OS independent manner. I would hope that the 1003.1 and P1003.1a interface definitions are compilable with C++ already. There seems to be a tendency in the postings here towards adding a lot of classes to the standard library. I'm concerned that we'll end up with more than are really appropriate and especially concerned that we'll get stuck with some that aren't as well thought out as they need to be and hence lack generality and portability to the world at large. -- ______________________________________________________________________________ Internet (vastly preferable) : rja@edison.CHO.GE.COM UUCP (if you've got no choice): ...uunet!virginia!edison!rja ------------------------------------------------------------------------------
mjv@objects.mv.com (Michael J. Vilot) (08/14/90)
In Article 112 rja@edison.cho.ge.com suggests that services which are more appropriately part of the POSIX IEEE 1003 standard should be left out of the X3J16 standard. This is the suggestion the library working group made to the X3J16 committee -- C++ bindings to other standards should be considered outside the scope of X3J16. One reason we are setting a priority on C++/C library compatibility is to provide access to the existing libraries and standards through already- developed C bindings. (Re-use at work ;-) Two other concerns were > that we'll end up with more than are really appropriate and ... > some that aren't as well thought out as they need to be I don't think the first issue is a problem, given the pace of progress of standards committees. (1/2 :-) The second issue is a valid concern. The goal of the library working group is to work from existing, proven libraries rather than to invent as part of the standards effort. Unfortunately, C++ experience with exceptions and templates (which both significantly impact library design) is rather scarce and not widely documented. I think both of these factors will work to limit the contents of the C++ standard library to elements that are widely regarded as necessary and helpful. -- Mike Vilot, ObjectWare Inc, Nashua NH mjv@objects.mv.com (UUCP: ...!decvax!zinn!objects!mjv)