The new scheduler works (almoast)

[Matti Aarnio <mea@nic.funet.fi>]

Hello,
	task to create the new scheduler is approaching its completition,
	but yet there are some loose ends to be plugged before it can
	truly be used in production environment.

	I have been testing it to see if the new way of starting up farms
	(or "rings") of transport agents works.

	They do work MOST of the time, but sometimes recipient addresses
	are left without processing just into the queue waiting for
	resources to become free, or some such.. (And all of the resources
	are in the idle pool waiting for jobs..)

	In spite of that problem, I was able to proof, that now the
	SMTP-clients can be started in processor-pool style manner,
	and while suitable threads exist in the recipient address rings,
	("ring" of addresses is such that the same transport agent process
	 can handle them all -- in sequential manner, of course.)
	the new scheduler is able to hop from one thread to another,
	and feed them to the transporter agents.

	With the new scheduler, memory leakage at the transport agents
	is likely to create memory bloat, because if all goes well, TA
	processes are started when they are needed the first time, and
	after that, they are around for a long time -- usually in idle
	state, though..

	We live interesting times :-)

	/Matti Aarnio <mea@nic.funet.fi> <mea@utu.fi>