I've been `using' the NoCeM support for a little while now, but it's
only recently that I got it to *work* -- that is, to save a cache
file.  It brought up some inconsistencies in the NoCeM code that I'd
like to bring up for discussion:

(This discussion relates to gnus 5.4.64, but it's known to occur in
the earlier 5.4 releases)

1.  gnus-nocem-scan-groups seems incapable of scanning for NoCeM in
    foreign groups.  That is, virtual group names in the NoCeM/active
    file have the undesired effect of scanning the local version of
    the group instead.

    This sounds like a stupid point, but my local news server doesn't
    carry all of the NoCeM bulletin groups.

2.  Are all of the NoCeM issuers spelled right?  It appears that
    "clewis@ferret.ocunix.on.ca" and "jem@xpat.com" may be spelled
    incorrectly in the gnus-nocem.el file.

3.  The real deficiency is in how gnus-nocem-enter-article decides
    when a message-id should enter the nocem cache.  Basically, unless
    one of the groups mentioned in the nocem article matches a
    locally-subscribed group, the scan will miss.

    I noticed this problem because I orthogonally set up my newsrc so
    that *all* groups are foreign...  I watched nocem run for about a
    week without matching a single article!

The way I've worked around points 1. and 3. above is to slightly
modify gnus-nocem-enter-article to match groups out of a *different*
hashtb than gnus-newsrc-hashtb.  This new hash table consists of the
gnus-group-real-name mappings of all of the groups in the newsrc.
This way, the nocem groups match up, and I can start accumulating
message-ids.

Please let me know if I've been approaching this in the wrong way!