> Now, when you link -- statically or dynamically -- you are including > portions of libc5 in your binary. This results in your binary being
Umm, no, actually -- the whole point of dynamic linking is that you're *not* including portions of libc5 in your binary. A replacement libc5 that met the "interface" of the one you used could be dropped in instead. (#including header files, that counts -- but not linking -- and it's sometimes surprising how much code can get away without using the header files...) The same is true of .dll's and *that* is the crux of the discussion. Now that I've been informed that libc5 is really under the LGPL (or at least parts of it claim to be) and that the /usr/doc/libc5/copyright file is *wrong*, I can certainly see a difference between that and cygwin32.dll. Nonetheless, neither is anything like QT. For some more perspective on the "interface" argument, go back and see some of the flaming a year or two ago about the GNU "libmp" (multiple precision integer math library.) See also the discussion of just a week or three ago about a company shipping a commercial package that uses GNU RCS underneath -- but since GNU RCS is built as a DLL (and they ship sources for those changes, and gnu rcs itself) they don't have to ship the program sources (and have allegedly run this past the FSF for confirmation that it was OK....) Recall that RCS is GPLed, not LGPLed. Isn't this fascinating? :-) I must admit that I'm glad to see, all in all, that this discussion has stayed *so* polite in comparison to the typical gnu.misc.discuss or other open net thread. Thanks! -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] .