Follow-up Comment #12, bug #18767 (project freeciv):

I don't know the details of the CMA algorithm, but I think Pekka's patch is
suggesting that "local" gold "usage" by the city is causing the "spurious" tax
collectors. I.e., if a city has lots of buildings/units consuming gold, it
feels like it needs to produce more gold. (Is that right?)

The patch makes the CMA no longer consider city usage[] at all when working
out targets for any O_GOLD, O_LUXURY, or O_SCIENCE.

While for O_GOLD (and O_SCIENCE) this kind of isn't true (but see below), so
the patch works, locally produced O_LUXURY *is* consumed locally, so I think
the CMA *would* need to take local luxury consumption into account.
* However, this doesn't matter for any standard ruleset, as there are no units
(or anything else) that *directly* consume "luxury" (as seen in city usage[])
-- rather, luxury is consumed to combat unhappiness, which is accounted for
more indirectly. Assuming that's dealt with properly, this problem with the
patch is largely theoretical. (Direct luxury consumption implies either
negative luxury or forced disband, so probably isn't sensible anyway.)

However, I think there's a wider issue, with whether such behaviour is a bug
at all.

The issue here is (I think) whether the CMA's economic calculation takes into
account what's going on outside its city.

gold_upkeep_style controls whether units/buildings' gold expenses are
accounted "to the nation" or "to the city", but in terms of game mechanics, I
think this only has any effect when the nation actually runs out of gold (when
it affects which assets are sold) -- otherwise, the city can always get a free
loan from the national treasury if it overspends, even chronically (and by
this means can be subsidised by the other cities), so I think there's no other
net effect of gold_upkeep_style on game mechanics.

However, I think gold_upkeep_style also affects the _behaviour_ of the AI and
of this city governor.
* If spending is accounted "on paper" to an individual city, the city governor
feels obliged to do their bit to help the national treasury (and to avoid
their assets being sold in case of crisis), ignoring the reality of wealth
** If all cities are governed by CMA, I think this tends to lead to a balanced
budget without player involvement (assuming no non-city spending)?
* However, if "on paper", spending is a national concern, local governors
don't feel it's their problem; they can spend as much as they like, and it's
the problem of the national government (i.e., the player!) to balance the
nation's budget (possibly by giving cities orders via CMA surplus targets).
Also, if the money runs out, whether the city covered its outgoings has no
effect on whether its assets will be sold, so there's no incentive for the
city governor to take it into account. (Yes, I know the governor isn't
_really_ thinking like that, but it's helpful as an economic argument.)

So this could be argued to be a feature, not a bug :)

However, I don't know whether this was an intended or unintended consequence
of gold_upkeep_style. The original change (r15610
<>, RT#40619
<>) doesn't provide much
in the way of rationale. Perhaps Matthias could comment?

A concrete way to test whether any of this is relevant would be to change the
ruleset to set gold_upkeep_style=2 (I assume the problem reported here is with
the default ruleset, where gold_upkeep_style=0? Is there a savefile?) and see
if the CMA's behaviour changes.

One possible hack to the CMA to make it take account of wealth redistribution
even when gold_upkeep_style=0/1 would be for it to peek at the national
treasury level; if times are good (lots of gold), it could ignore
usage[O_GOLD] and assume it'll be bailed out; but if the national treasury
runs low (defined somehow), it could start taking its own spending into
account to preserve its own units/buildings (probably meaning less output of
other kinds -- um, simulating liquidity crisis conditions?). Do we want the
CMA trying to be that clever?

(Anyway. All of this started out as a quick look to see if Pekka's patch could
be quickly applied for 2.3.2. I've decided not to, for now.)


Reply to this item at:


  Message sent via/by Gna!

Freeciv-dev mailing list

Reply via email to