there seems to be a lot of questions about how to select a cylinder head for a given application. I'll give you an idea of how/why i choose a given head for a given application. One thing to keep in mind is that nothing is ever perfect. People who have dedicated there entire lives to cylinder head developement still change there mind daily about what is best and most important. The information presented is mostly gathered from peers in the industry and can be used as a general guideline. As a whole these numbers are generally agreed upon, but these are not rules written in stone.
First things i want to know are these-
budget
cubic inch
budget
rpm
budget
intended application
and did i mention how much you have to spend?
It will never cease to amaze me how many people want an 8000 rpm small block and have an old set of "camel humps" and $250 to spend. Some things just aren't doable.
When i have that information there are some formula's that i use to help me determine where i want to start. I have a pretty good data base that allows me to choose a head based on the cross sectional area available. This minimum cross section will determine when the head will "shut off" or fail to make decent power past a certain rpm level. I'll use a typical engine for an example. We'll build a hypothetical 500hp 350 engine. We'll use 7200 rpm as our target rpm. We will also use a standard 4.00x3.48 bore/stroke. Using this rpm level and bore/stroke this is the formula i would use to get a baseline minimum cross section needed-

(bore x bore x stroke x rpm x .00353)/614

Using this formula and our numbers we can determine that we need a min. cross-sec of about 2.30 square inches. This will allow our motor to turn 7200 rpm without exceding 614fps or .55 mach (the same thing). That number is generally considered to be the point at which most "conventional" type cylinder heads will reach a point of choke. A modern pro-stock style head moves that number up a bit, because of port efficency. A flat head wouldn't even come close to that number. If you put numbers in for a typical 400" small block you would see that it would take about 2.64"sq to make power to the same rpm level.Quite a bit larger. If you go the other direction, that same head with 2.3" on a 302" engine would put your rpm level up around 8300 rpm! Makes it easy to see why a smaller motor will "rpm", huh? On a conventional aftermarket small block chev head, the pushrod pinch area represents the smallest cross section. Normally 1.050" is about all you will safely get at the width of the port. This means we need about 2.2" of height in the same spot. This would probably put you into the 210-220cc volume for a typical head. Now it starts to get a bit tougher. When you are porting a cylinder head localized velocities are what will make or break a head. When i say localized velocities, i mean the actual measured air speeds when flowing it on a bench. These are not to be confused with the air speeds generated by rpm/cubic inch/cross-sec. that we just talked about. These are numbers that can be figured using a formula with measured airflow and cross-sec or by using a pitot. A pitot is a steel tube with a small hole in it that will measure pressure differential when put into the airstream. They look like this-
http://www.superflow.com/flowbenches/index_1017.cfm
We use these to figure out how fast the air is moving through various parts of the port. I'll take and stick the pitot in the port and measure speed at the top of the short turn, across it in three or four different spots and three or four different levels. I'll take other measurements in the same manner at different spots including the pushrod pinch, and opening of the port. I prefer to use this method of measurement to the "calculated" method because it is hard to get an accurate measurement of the cross section at the short turn, and in other spots, without making a port mold, cutting it and measuring it. Even then using the calculated method you will only get an average, not the true localized velocities. In most conventional heads, you don't want these velocities to excede 350fps, as measured. So if i probe the port and find that at .500 lift my short turn has speeds that are 450fps, i need to do something to that port to change that. Anything that is over my 350fps limit is going to create turbulence and separation of the air in the port. This is going to limit my ability to fill the cylinder as well as i can. This is where the time and energy spent by a good head porter separates themselves from the fluff and buff crowd. Knowing what to do in these instances is what makes a good head. The formula used to figure out flow bench air speed based on flow in cfm @ 28"h2o and cross sec. is this-

(cfm/cross sec) x 2.4
this will give you air speed at that point. If we use our example of 2.3" at our minimum cross sec you'll see that we would require about 335cfm @ 28"h20 before our head would have a flow problem at that point, at least based on our numbers anyway. This is why going in there and wacking that area larger isn't always the best thing to do and why having a huge opening at the gasket doesn't do anything but slow that air speed down. We want to approach that 350fps as best we can, to get maximum filling, without stepping over it.
I just "fixed" a set of heads on an engine. It had a problem with short turn airspeeds. The head flow well for what it was, but the short turn needed some work. I spent about 10 minutes in each port fixing the area. The head gained about 2 cfm of airflow. On the dyno it picked up 27hp and extended the usable power range of the head by about 400rpm. All with 1 1/2 hours worth of work. Pretty good gain for just paying attention to one small thing, huh?
If you pay attention to air speeds, localized velocities, and cross sectional areas you will end up with a head that is driveable and makes good power. I don't even flow a head for a "curve" until i have satisfied these needs. You need to be realistic with power expectations and rpm levels when you set out on your project. An 8000 rpm n/a 327" motor probably isn't going to be very streetable. Almost allways error to the side of small and you normally won't go wrong.