On Aug 19, 2008, at 6:24 PM, Greg A. Woods; Planix, Inc. wrote:
On 19-Aug-08, at 7:03 PM, James Chacon wrote:On Aug 19, 2008, at 5:47 PM, Aaron J. Grier wrote:On Tue, Aug 19, 2008 at 03:32:38PM -0400, Greg A. Woods; Planix, Inc. wrote:On 19-Aug-08, at 3:13 PM, Aaron J. Grier wrote:again with the binary-only strawman. I do not understand how you getfrom simplifying the build process by eliminating .a files to in marketing perceptions and proprietary OS vendors. perhaps you could elucidate. let me see if I can summarize your argument: - user must be binary-only OS consumer AND - user must be source application consumer AND - user must be interested in statically-linking source applicationsANDuser must be able to compile his own source but somehow cannot possibly compile/obtain NetBSD sourceOtherwise in Greg's world the user would simply be able to compile both ends of whatever this user wants and have no issue at all.Obviously you guys haven't spent much time working around GNU/Linux users lately. (or open solaris users for that matter)Some folks will be willing and able to blindly type "configure && make install" and some folks will even write their own code, but not all those folks, in fact the vast majority in both those camps, will be willing and able to rebuild their entire OS from scratch with the custom options necessary to get static libraries installed, _even_ if it is NetBSD and it continues to be as easy to build as it is today. As others have clearly stated, some of those users will want to at least try static linking sometimes, even if they don't always understand why.
1. Show me a user there who is so inept that all they can do is "configute && make install" who actually cares about static vs dynamic linkage. I'm willing it wager a fair amount you can't find any of these users...
2. Who said anything about rebuilding the entire system? If there's a target saying "build static libs" and that's all someone has to do to get them (for the 1-2 people who might actually be affected who aren't otherwise building their own systems, so folks like you/mouse/etc don't count...).
Meanwhile what you guys are foisting is the real straw-man argument. Why stop shipping static libraries when they can be, and indeed probably still _must_ be, built during a system build? Why use such stupid and lame arguments such as PAM (and maybe CITRUS) to bolster this proposal? Why ignore all the real issues with _forced_ dynamic runtimes? Nobody has managed to give any real technical evaluation showing any real, i.e. as opposed to straw-man, arguments in favour of _forced_ dynamic runtime.
Really...We need .o's to build the system, we need all sorts of intermediate output...We don't ship it either because it's not needed for general use. Same way with static libs there. Nothing needs them by default...
James