Recognizing someone is a tough issue.  One way is to have everyone unidenitified until introduced.  The character keeps a memory list for each character it knows.  However, it can get quite long if you are meeting new characters and npcs over months.  If you cap this, you run into problems of forgeting people.  (This is not entirely bad and could be based on intelligence stat.)

As for talking to npcs, the goal is for them to listen when talked to but not respond to meaningless chatter.  But sometimes that is fun.  I once was playing around in a MUD I was coding for.  I decided to create a secretary npc that would show people into my quarters and give messages when I was out.  (Sometimes boredom is the mother of invention.)  I got her detailed enough to respond to so man words, that I renamed her to a generic player name and stuck her in the middle of a town.

I watched people interact with her and got a kick as how often people thought she was a real person.  It would take awhile for them to figure out that she was not a pc.  But it worked because she listened to everything and would respond to certain keywords.

So I would suggest having some nocs that listen to everything as well as some that only listen when activated.

Quath

I like the philosophy of the change.  I like the /greet idea.  By default it could say "Hello, $targetname."  But it could be extended to take arguments for customization like

/greet em tips his hat at $targetname.
/greet say Well met, friend.
/greet shout Great to see you again, $targetname!

They could be macroed and used when appropriate.  This would allow for an npc to be woken up in a roleplaying manner suitable for most everyone.

Quath

As I trecked another bundle of rat skins along the road, I had some time to think about the economics of the situation.  Who uses rat hides?  Do they make good pouches?  Or clothes?  Or paint brushes?  The gushy eye balls were another commodity I was peddling but was worried about finding out how they were really being used.  Maybe potions?  Or stews?

But more fundamentally, with all the rat parts being sold, why were vendors still paying so well for them?  With such a huge surplus, the price should fall so people will find other things to do for money.  In other words, a more dynamic economy.  I have seen a few ideas so far on this here, so I figured I would offer mine as well.  It starts off with a slightly different problem...

How many NPCs live in a city? - Is it a hundred?  Thousand?  Few thousand?  You can't have them moving around at once or it would be too confusing.  But it would be nice to be able to deal with a city and its population.  So here is my simple idea for a town/city.  Set the following information for each town/city:

• Population
• % of population is human (and store racial percentage for each race)
• % of population are artists, soldiers, smiths, healers, etc
• Index of locations in town. 
• random number seed

Generate a npc - So for fun, you could generate a random person.  Say if the town had 100 people, you could use a random number to get 1 to 100.  That number (say it is 23 for this example) is added to the random seed to be used to generate that person.  The random seed picks a name from the name generator, a sex (50%/50%), a race (from the city information), a profession (from city information), and body type.  This information should always be the exact same for the 23rd person of this town.  From there you can sample a starting location (home) and several points of interest to go to and then back home.

This adds some interesting atmosphere as you recognize npcs and see them go about their day.  It also adds some persistency because you can figure out where each person lives and it will stay the same (until the percentages are changed).

Economy - Full blown capitalism is hard to model.  After all, if rat handbags became too expensive, people may buy silk handbags or burlap or just carry things by hand.  It is hard to model full supply/demand.  But I think there can be some fun approximations to it.  So for each item, a target price could be set.  Also its desire (and desire of goods made from it) by profession could be entered.  I see exporters or caravans as a profession for this concept.  So for example a rat hide:

Target value - 20 tria.
Artists 2, coat makers 50, smiths 20, farmers 30, exporters 100 (numbers represent how many rat hides a 100 people of that profession would buy in a real life day)

So say there is a city with 200 artists, 100 farmers and 10 exporters.  They would want to buy 4 + 30 + 10 = 44 rat hides a day.

So we can figure out the surplus for each city:

surplus = number hides gathered by pcs + number gathered by npcs - number bought by pcs - number bought by npcs.

The only thing unknown is how many npcs are farming hides.  This would have to be entered on a city by city basis.  The default could be 0 for a player driven economy and = number bought by npcs for a naturally balanced economy.  But once that is known, the surplus can be counted.

The price can be adjusted as

new value = target value * (1 - surplus / number items bought)   Min value = 1, Max value = 10 * target value

Issues
1. Lots more information has to be entered.  However, if you set up the city information to generate random people, the a lot of it has already been set.  Each item would need to define its desire by profession and how many producers of this good per city (or a good default).

2. Need to store bought/sold information for each item for each city.  This is not too bad, but costs more as number of istems and cities grow.

3. Computing new prices could be expensive.  My suggestion would be to run a script at night (every 24 hours) to compute the new prices for the next real life day.

4. Lots more room to customize cities.  There could be a plague or a war that decimates a population.  The economy could adjust until the population rises back up.  More exporters could be created to equalize prices among cities.  Or fewer if the city is far away.