Re: [Axiom-developer] Axiom silver branch

[root]

> Storing binary files in a source code archive is something that
> I have always objected to, but Tim insists on re-distributing
> code this way. I really think we should stop doing this.

Binary files exist no matter what we do. The Axiom root directory
contains binary image files as well as the src/doc/ps subdirectory.

As for the 'redistributing code' we differ on philosophy. 
I've written extensively on the difference in philosophy.

Bill and several others have taken the stance which can be positively
characterized as 

"first, we assume that the users have all the prerequisites...."

and negatively characterized as

"the users will have to guess...."


On the other hand I've taken the stance that can be positively 
characterized as

"axiom builds should 'just work'..."

and negatively characterized as

"fork the world and do it ourselves..."




I am regularly fielding questions of the form:

   > Camm,
   > 
   > I tried to build Axiom on a debian machine and get:
   > 
   > creating makedefc
   > sed: file conftest.s1 line 61:  unterminated `s' command
   > creating windows/gcl.iss
   > sed: file conftest.s1 line 61:  unterminated `s' command
   > creating windows/sysdir.bat
   > sed: file conftest.s1 line 61:  unterminated `s' command
   > creating windows/install.lsp
   > sed: file conftest.s1 line 61:  unterminated `s' command
   > 
   > 
   > suggestions?
   > 
   > Tim

   Yes - my guess -- emacs is not installed?  The Debian package is setup
   to require the packages listed under Build-depends: in debian/control
   at build time, one of which is an emacs.  The error message obviously
   needs work :-).

In a system as complex as Axiom it is impractical to expect users to
understand all the dependent paths. Even I get confused. However when
I understand the confusion I write makefile changes that ensure that
others do not have the same problems.

We COULD distribute Axiom and watch the builds fail over and over
again because users have broken versions of noweb or their native
version of GCL does not include the correct libraries or emacs is
not installed or their version of some software is not right for Axiom.

The practical effect of such a policy is that every user has the
responsibility of upgrading their system to meet our requirements.
And it means that Axiom WILL NOT build successfully on most systems.

I suggest you use the 'silver' system to experiment with trying to
remove binaries and require native GCL/noweb installs. My experience
tells me that you'll find it won't build everywhere. It certainly
will NOT build on any of my current systems as my native installs
do not include the required axiom patches.


Alternatively we COULD pursue the philosophy that Axiom should 'just work'.
That means that Axiom WILL build successfully on most systems. The 
practical effect of such a philosophy means that we distribute binaries
with patches until the patches are accepted in the upstream systems AND
the upstream systems become the standard in most distributions.

I still hold the fundamental philosophy that 'Axiom should just work'.

Tim