Actually I'd change that slightly:
1) Filter/sanitize input (php.net/filter works great for this)
2) Sanity check input (bounds/type checking and then check if it is
legit against the datasource)
a) I use intval() when I expect numeric input. It will scrub out
anything that isn't an integer. Depending on your needs, additional
functionality like a regexp or substr() might be needed but that's up
to you. Remember to get it into the right type first.
b) I always include a default value in case the value does not exist
from the input too.
3) Store data in its original form, do encoding/other things at
runtime on output
4) Encode output (so the browser is not gven <script> tags but
<script> for untrusted content for example)
When dealing with db queries, use mysql_escape_string or equivalent at
runtime, prior to the query. Makes it easier to keep track of things
that way, instead of escaping the data up top.
example:
# sanitizing, filtering
$foo = filter_input(INPUT_GET, 'foo', FILTER_SANITIZE_STRING);
# if default value is needed
if(empty($foo)) {
$foo = 'default value';
}
# escaping - could make this into a single line too by using sprintf
or other things.
$foo = mysql_escape_string($foo);
$result = db_query("SELECT foo FROM bar WHERE baz='$foo'");
for numeric input:
# sanitizing, filtering - you'll get a type int with anything
malicious scrubbed out already
$foo = intval(filter_input(INPUT_GET, 'foo', FILTER_SANITIZE_NUMBER_INT));
# bounds checking examples (assume $maxpages has already been determined)
if($foo < 0 || $foo > $maxpages) {
redirect user back to foo.php?foo=1
}
(pretend $foo is the page number. I could have made a better example there)