Some design features (as I see them):
The basic mechanic is very simple: d20 + modifiers vs. spell DC + modifiers.
The modifiers allow a caster more flexibility without needing feats; silent spell and still spell now simply increase the difficulty, and material components are always optional (but a good idea--in playtesting, my caster has been choking on how quickly his vitality goes down).
The feat progression then becomes a choice between power (concentrating on one school and becoming better and better at it) and flexiblity (using feats to gain access to other schools at the expense of power). Once again, nothing too unusual.
The real nuts and bolts, however, come from the design of the skills. After playtesting earlier versions, I realized that I would have to create some weights for the different effects, because all effects would never be equal. I also had problems in earlier versions with some skills being seen as almost useless, while others were almost universally must-have.
First, I expanded or contracted some of the skills. Evocation proved still too powerful, so I broke the feat up into the four basic elements (after experimenting with less elegant solutions such as doubling the skill points per rank needed) and broke up what had been one huge skill per element into three.
All of the effects are now weighted in terms of difficulty and power. I've tried to have similar effects have similar difficulty and power (although I certainly may have slipped up somewhere--typing out chart after chart gets mind numbing after a while). I've taken some magical effects that are staples of normal D&D magic, but tend to run contrary to the desired style of some low-magic DMs (such as fly and teleport) and made them more difficult. If you want a different flavor with the system, it is relatively easy to adjust difficulty and power.
Another way to adjust the power levels is to decide how common magic items and material components will be. These spells suck vitality from low-level casters at a prodigious rate. If you wanted greater power, you could have items that worked like psionic crystals, or wands that stored vitality that only worked for a particular skill, or various focus items that gave a benefit to the spell roll (allowing the caster to more easily modify the spell, especially the saving throws--one trade off I have intended is that, in exchange for forcing a spell roll, it is easier for a caster to increase the save DC against a spell).
I have two futher limiting factors on magic that, as far as the design is concerned, are optional, but do have the effect of making magic rarer. The first is "arcane awareness," which allows magically tuned individuals to possess a kind of magical radar to be aware when spells go off. The caster in our group is very aware of this, and he always seeks non-magical solutions first so that he won't tip off any other casters in the area. It also allows him to be aware of other casters, which gives the caster a usefulness even when not casting spells.
The other limiting factor is corruption, which can come from channeling more vitality per round than a caster can safely handle. This mechanic is intended to lengthen casting times, but offer the temptation for when things need to happen quickly. If you don't like it, you can just make the vitality-per-round a hard limit.
The other use of corruption is more flavor-oriented. In this conception, magic is create by living beings, so turning that power against living beings in a very direct way creates a corrupting effect. I limited this to: taking someone's life or dominating someone's will. I recognize that these are both fuzzy standards; judge for yourself what might corrupt someone, or ditch this altogether. The game effect this mechanic has is discouraging what many consider to be "flashier" magic (the fireballs and mind-control spells) in favor of buffing and reconaissance, without preventing mages from having the ability to destroy things when they have to. And the checks are low enough that, if you work to keep your corruption low, you shouldn't have problems when you really need it. My own player is in a love it-hate it relationship with this rule: he hates the limitation, but loves the role-playing opportunity that this presents.
The system is not finished. Mainly, I got tired of writing up and working through various skills. The most glaring omissions to my mind are a lack of phantasms (spells that combine divination and illusion to create images drawn from the target's own mind, and therefore more effective), animation (transmutation--useful for constructs and golems), and some spell series that creates pocket dimensions (conjuration, for rope trick, magnificent mansion, portable holes, what have you). I'm sure there are others that haven't occurred to me yet.
Anyway, those are some of the thoughts that informed this design. I appreciate any feedback these boards might have, and will be happy to try to answer any questions.