Possibly I'm just unusual in how I set things up, and no-one else would use
this.  I'm just throwing this suggestion out there on the off-chance...

Currently my incoming mail is checked against a whitelist, and only mail from
unknown senders is passed through spamassassin (with bayes autolearn).

However, almost all my ham hits the whitelist, and hence isn't run through
spamassassin. So in order to provide a decent ham corpus to the baysian
classifier, I pass all whitelisted mail through sa-learn.   This probably also
gives the classifier a much better ham corpus than it would get through
autolearn, although the original motivation was efficiency on an underpowered
machine.

I've just looked at switching to spamc/spamd, and I realize that whilst I can
replace all my calls to spamassassin with calls to spamc, I can't do the same
with my calls to sa-learn.

So my RFE is for a flag(s) to spamc to cause sa-learn behaviour, ie to request
that spamd run the baysian learning algorithm on the message.

Does this sound sensible?  Is it likely to be useful to anyone other than me? 

-roy