Date: Wed, 28 Aug 1996 15:54:57 -0500 From: rkw@shark.dataplex.net (Richard Wackerbarth) To: Nate Williams <nate@mt.sri.com> Cc: hackers@freebsd.org Subject: Re: Am I wrong or is this just stupid?r Message-ID: <v02140b01ae49b2cb82c0@[208.2.87.4]>
next in thread | raw e-mail | index | archive | help
The comment about profit was meant to be humor.
The problem with "creating a working model" is that it takes the entire
system to demonstrate that it works.
The problem with "has to work at least as well" is that this is too
subjective. Every idea that changes ANYTHING has someone who complains that
you broke their favorite "feature". We have to decide just which are the
sacred cows.
The problem with failing to have a design adopted before it is implemented
is that you are shooting at an impossibly moving target. As an example,
take the suggestion of Toshihiro Kanda <candy@fct.kgc.co.jp> which just
happened to be in my in box. (No reflection on the merit of his idea)
[...]
+ if (shadow_directory) {
+ #define SYSDIR "/usr/src/sys"
+ (void) strcpy(xxx, SYSDIR);
+ if (symlink(xxx, path("@"))) {
+ perror(path("@"));
+ exit(2);
+ }
+ (void) sprintf(xxx, "@/%s/include", machinename);
+ if (symlink(xxx, path("machine"))) {
+ perror(path("machine"));
+ exit(2);
+ }
+ #undef SYSDIR
+ }
+ else {
(void) sprintf(xxx, "../../%s/include", machinename);
(void) symlink(xxx, path("machine"));
[...]
Such a change would be counter to the direction that I am proposing. It
imbeds absolute paths in yet another place. Unless "we" decide what
direction we are going, others will continue to work against my goal faster
than I can make progress. However, if we do have a goal and a plan to get
there, I am sure that they would be more than willing to make their changes
to help make things conform to that plan rather than subvert it.
Here is my plan:
1) Re: Make world
We should "unroll" this. As others have stated, the necessity to
recompile the compiler is overkill in our releatively stable environment.
Since I propose to divorce the entire make process from the underlying OS
environment, those who feel that they need to do the "bootstrap" operation
can manually do so. For the rest of us, it is simple
cd <top_of_tree>
The really brave can
ln -s / root
Personally, I would be more inclined to
mount ..... root
Then you simply
make clean
make all
make install
and for Jordan, et al
mkdir /pub/FreeBSD/960828-SNAP
ln -s /pub/FreeBSD/960828-SNAP root/usr/distribution
make tarballs floppies
2) I would propose the following structure to <top_of_tree>
bin/
inc/
lib/
src/
obj/
root/
root/bin
root/usr
root/usr/bin
[etc]
Tools (all the commands referenced in the makefiles) would be placed in "bin"
Making a library would 'install' it in "lib" and its headers in "inc".
This is where the files would be found when making the userland commands.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?v02140b01ae49b2cb82c0>