+Files we use and update
+-----------------------
+
+The program reads and writes the following files:
+
+ * #pages#.pnm
+
+ Contains one or more images (as raw ppms, end-to-end) which are the
+ screenshots taken in the last run. This is (over)written whenever
+ we take screenshots from the YPP client. You can reprocess an
+ existing set of screenshots with the --same (aka --analyse-only)
+ option; in that case we just read the screenshots file.
+
+ You can specify a different file with --screenshot-file.
+
+ If you want to display the contents of this file, `display' can do
+ it. Don't try `display vid:#pages#.pnm' as this will consume
+ truly stupendous quantities of RAM - it wedged my laptop.
+
+ * #master-newcommods#.txt #local-newcommods#.txt
+
+ Dictionary of newly introduced commodities. When a new commodity
+ appears in Puzzle Pirates, the PCTB server operators need to add it
+ to their database for us to be able to upload data about it.
+
+ It can sometimes take a few days to do this. In the meantime, it
+ is possible to upload partial data - data just omitting that
+ commodity. This is controlled by these files: they list
+ commodities which should be automatically ignored if the PCTB
+ server doesn't know about them. The master file is downloaded and
+ updated automatically from my server. You may create the local
+ file yourself. The format is simple: one commodity per line.
+
+ Unrecognised commodities can also be due to OCR failure so
+ double-check what you're doing before overriding the uploader by
+ telling it to ignore an unrecognised commodity.
+
+ * #master-reject#.txt #local-reject#.txt
+
+ Dictionary of regexps which, when the OCR appears to match, we
+ reject instead. At the moment this is used to stop us thinking
+ that `Butterfly weed' is `Butterflyweed'. This happens if the
+ character set dictionary is missing the lowercase `y ' glyph.
+ See README.charset.
+
+ * #master-char*#.txt #local-char*#.txt
+ #master-pixmap#.txt #local-pixmap#.txt
+
+ Character set and image dictionaries. For the semantics of the
+ char* files README.charset. There is not currently any accurate
+ documentation of this dictionary format.
+
+ #master-*#.txt contain the centrally defined and approved data.
+ They are downloaded automatically from the SC PCTB server and
+ updated each run. You can safely delete this file, if everything
+ is online, if you want to fetch a fresh copy.
+
+ #local-*#.txt are a local copy of your submissions, so that they
+ will be used by your client pending approval by me. You can delete
+ this file if you think you may have made a mistake.
+
+ See README.privacy for details of the communications with the SC
+ server about the contents of these dictionaries.
+
+ * #commodmap#.tsv
+
+ Map from commodity names to the numbers required by the PCTB
+ server. This is fetched and updated automatically as necessary.
+ It can safely be deleted as it will then be refetched.
+
+ * #upload-1#.html #upload-2#.html
+
+ We screenscrape the pages from the PCTB upload server. The actual
+ HTML returned from the upload server is left in these dropping
+ files for debugging etc.
+
+ * <file>.tmp
+
+ When any of these tools overwrite one of the persistent dictionary
+ files, they temporarily write to <file>.tmp.
+
+Future versions may have more helpers and more data files.