[net.mail] definitions for LOCAL, DEMAND, DAILY

pedz@smu.UUCP (01/25/85)

The first problem is that pathalias assumes the cost of a link
is the same in both directions (at least I think it does).  This
is often not correct since site A may call site B on demand but
site B calls site A only once a day.  Thus if no requests are
made at A to go to B, a request going from B to A will have a
link of once a day.  I realize that when uucp calles up,
messages are passed both ways.  However, the weights are
basically describing the frequency of the calls.

The second point is that pathalias and its algorithms are only
the first attempt at solving the problem.  Therefor the site
description should contain as much as possible (within reason of
course) so that future programs may take advantage of it.  Thus
the "double" link problem (of having a site call another site on
demand and once per day) should be solved by putting the
information into the site's description and pathalias can ignore
it if it wants to.

Perry
convex!smu!pedz
pedz@smu    (csnet)

guy@rlgvax.UUCP (Guy Harris) (01/28/85)

> The first problem is that pathalias assumes the cost of a link
> is the same in both directions (at least I think it does).  This
> is often not correct since site A may call site B on demand but
> site B calls site A only once a day.

Wouldn't

	A	B(DEMAND)
	B	A(DAILY)

do here?  From the PATHALIAS(1) manual:

	"Pathalias" expects as input a sequence of host-to-host
	connectivity information, with a host name in column 1,
	followed by white space, followed by a comma-separated
	list of links (also host names), denoting a connection
	from the host to the links.  *Connections are assumed to
	be unidirectional.* ("italics" mine)

So the first line declares that messages are passed from A to B on
demand, but says nothing about the passage of messages from B to A.
The second line says that messages are passed from B to A at least once
a day.  (The fact that messages will be passed from B to A as soon as
a message is shipped from A to B should not be taken into consideration
in determining path costs, since this is not normally under B's control.)

	Guy Harris
	{seismo,ihnp4,allegra}!rlgvax!guy