Can somebody tell me what FM to R for this?

I've been internationalizing a game database I've had for some time.
Things were working reasonably well today, and then all of a sudden
everything went away -- every attempt to pull a web page caused
a failure in the call to gettext.translation.  Eeeek.

It turned out that I had just installed a file in
    /usr/share/locale/en/LC_MESSAGES/ohex.mo
and for some reason this completely murdered any attempt to
establish a non "en" translation.  Even figuring this out took some
sleuthing, because the error trace is completely unhelpful:  what
is one to make of
    LookupError: unknown encoding: CHARSET
when trying to use a previously-working translation?

Anyway, what I'm really trying to do is to create a "null" translation
that just shows where the literals in the python code actually
are used in the output.  The code is complicated enough, that
what I really want is a non-translated version.  I think this will
help my translators figure out which line of their *.po file fits into
what context.

I tried sticking it into the C tree, but this did nothing at all.  So
I thought I'd stick it into the vanilla "en" tree, but this made
_everything_ fail.

My Python in a Nutshell is no help.

So:  a bit of help of the immediate topic would be nice.  Perhaps
better would be a pointer to a useful (not too bloody theoretical
or abstract, please) FM to read.  I don't want to read any BNF,
I just want practical help.

++ kevin