diff --git a/www/docs/patch.t b/www/docs/patch.t
new file mode 100644
index 0000000000..8c601b0b14
--- /dev/null
+++ b/www/docs/patch.t
@@ -0,0 +1,58 @@
+#define _PAGE_ How To Work With Patches
+#include "head.t"
+
+<h2>Tools Of The Trade</h2>
+<p>
+Use the tools 'diff' and 'patch'. Preferably the GNU versions. They're readily
+available for all imaginable platforms.
+<p>
+[Add favourite diff/patch links here]
+
+<h2>Creating A Patch</h2>
+<p>
+ We generate diffs (often called patches) using 'diff' in a manner similar to
+this:
+<pre>
+  diff -u oldfile newfile > patch
+</pre>
+<p>
+ People who have checked out code with CVS can do diffs using cvs like this:
+<pre>
+  cvs diff -u file > patch
+</pre>
+<p>
+ 'diff' can also be used on a whole directory etc to generate one file with
+changes done to multiple:
+<pre>
+  diff -u olddir newdir > patch 
+</pre>
+<p>
+ The -u option means the output is using the 'unified diff' format. Older
+ diff programs don't have that, and then -c (for 'context diff') is OK.
+
+<h2>Applying A Patch</h2>
+<p>
+ Applying a 'patch' (output from diff -u) is done with the 'patch' tool:
+<pre>
+  patch < patchfile
+</pre>
+<p>
+ patch knows that the patchfile is a set of changes on one or more files, and
+will do those to your local files. If your files have changed too much for the
+patch to work, it will save the sections of the patch that aren't possible to
+apply in a file called "filename.rej" (filename being the name of the file for
+which the failing section was intended for). Then you must take care of them
+manually.
+
+<p>
+ If there is path information in the patchfile that you want to cut off
+ from the left, tell patch how many directory levels to cut off to find the
+ names in your file system:
+<pre>
+  patch -p0 < patchfile
+  patch -p1 < patchfile
+  patch -p2 < patchfile
+</pre>
+ ... each example line removes one extra level of dir info from the left.
+
+#include "foot.t"