R help - Dec 2004 - Package dev: Depends, require, SaveImage best practices?

I'm trying to sort out some best practices for package development and
understand some behavior of R CMD check that has me confused.

Best practice question:  If a package foo appears in the Depends field in
the DESCRIPTION file of mypkg, should I refrain from using require(foo)
in the R source files of mypkg?

The Writing R Extensions manual says this about the Depends field:

    ... the list of packages which is specified will be attached before
    the current package, both when library is called and when saving an
    image or preparing for lazy-loading.

I conclude that I need not require() a package that I've listed in
Depends.  However, this isn't always working for me.  I have a package
that has Ruuid in Depends.  If I also have SaveImage: Yes, then R CMD
check completes without error.  However, if I omit SaveImage, then I get
this:

    $ R CMD check graph
    WARNING: ignoring environment value of R_HOME
    * checking for working latex ... OK
    * using log directory
    * '/home/sfalcon/proj-svn-unix/graph-proj/graph.Rcheck'
    * checking for file 'graph/DESCRIPTION' ... OK
    * checking if this is a source package ... OK

    * Installing *source* package 'graph' ...
    ** libs
    make: `graph.so' is up to date.
    ** R
    ** data
    ** inst
    ** preparing package for lazy loading
    Error in assign("nullgraphID", getuuid()) :
            couldn't find function "getuuid"
    Execution halted
    ERROR: lazy loading failed for package 'graph'
    ERROR
    Installation failed.

To conclude: what am I missing?  Would it be better to just
require(Ruuid)?  Always use SaveImage? 

Much tanks,

+ seth