Because installing software in Linux ranges from difficult to
impossible we've split up the distributions into three levels of
difficulty:<P>

<BLOCKQUOTE>

Development trees:  virtually impossible to compile without software
development experience but allow you to submit changes to the source
code.  These are hosted in <A
HREF="http://heroines.sourceforge.net">heroines.sourceforget.net</A>.<P>

Source code distribution:  very hard to compile from scratch but doable
in a single night.  It's very hard to submit changes to the source
code.  These are hosted in <A
HREF="http://heroinewarrior.com">Heroinewarrior.com</A><P>

Binary distribution:  run rpm and forget about it.  These are hosted on
<A HREF="http://heroinewarrior.com">Heroinewarrior.com</A><P>
</BLOCKQUOTE>

The current state of heroines.sourceforge.net is the raw development
trees.  There are scripts in the development trees which automatically
package it into source code and binary distributions which users can
download and install in less than a month.<P>

The reason for this system is 

<BLOCKQUOTE>

#1 The libraries and tools are very esoteric and rapidly evolving.  It
would be impossible for users to manually search and install every
library and tool required to install Broadcast 2000.  Instead, as many
libraries and tools as possible are statically built into the
executables, isolating the user from the bird's nest of standard lib
directories and Google searches.<P>

#2 The libraries and tools are shared across many applications. 
Changes to the libraries must be propagated to the applications without
involving the hairball of standard lib directories and without cp
-rd'ing the libraries after every change.<P>

#3 Some of the applications can't be distributed in the same tarball as
the libraries because of patents and licenses.<P>

</BLOCKQUOTE>

Enter the <I>symbolic linking</I> paradigm.  The biggest reason users
can't understand our development tree structure is that Windows is king
and symbolic linking doesn't exist in Windows.<P>

When you symbolically link something, it appears in the directory just
like the original file would, except it's really an illusion.  Changes
made to the original file appear in every symbolic link to it, as if
you changed many copies of the file.<P>

Here's the development tree you need to re-enact on your computer to do
development:<P>

<PRE>
./
./libmpeg3
./quicktime
./cinelerra
./xmovie
</PRE>

The first step in constructing this development tree is downloading
directories from CVS.  Sourceforge gives a <A
HREF="http://sourceforge.net/cvs/?group_id=13554">CVS command set</A>. 
Here's a CVS command set customized for Heroine Virtual:<P>

<PRE>
export CVSROOT=anonymous@cvs.heroines.sourceforge.net:/cvsroot/heroines
cvs co libmpeg3
cvs co quicktime
cvs co cinelerra
cvs co xmovie
</PRE>

The next step is to insert the symbolic links.  CVS unfortunately
deletes symbolic links so once you have a CVS tree you need to download
a sourcecode distribution from <A
HREF="http://heroinewarrior.com">Heroinewarrior.com</A> and compare the
trees to get the symbolic links.<P>

Don't expect to compile anything useful out of this.  When you compile
this tree you get executables which run where they are for testing
purposes.<P>

You compile the tree by running make in each of the four directories.<P>

Enter <P> 

<PRE>
export LD_LIBRARY_PATH=.:$LD_LIBRARY_PATH<P> 
</PRE>


into your
.bashrc, .bash_login files and log into another Xterm.  Now you can cd
into a directory and run an executable.<P>

The executables are currently:<P>

<PRE>
./cinelerra/bcast/bcast2000<BR>
./cinelerra/cinelerra/cinelerra<BR>
./xmovie/xmovie/xmovie
</PRE>

The plugins are currently:<P>
<PRE>
./cinelerra/plugins<BR>
</PRE>