Unless you really want to get into making an oscillator, don't. Just buy one.
If you are going to make an oscillator, use an HCMOS gate, preferably an unbuffered one such as a 74HCU04.
The frequency stability of an oscillator is not something that you should worry about. Unless you are looking at temperature compensated or oven stabilised designs, the frequency of the crystal will change so much with temperature that you will never notice the oscillator.
Also, if you don't have any way of checking the initial adjustment of an oscillator, it could be a long way off.
But initial adjustment is not important unless there is some reason for good accuracy.
Let me put some figures on that:-
Initial adjustment if component values are guessed. ±500 ppm or 40 seconds per day
Initial adjustment if component values are selected for the design. ±50 ppm or 4 seconds per day
Initial adjustment if component values are selected for each oscillator, or a trimmer is fitted. ±10 ppm or about 1 second per day
Temperature stability with a cheap crystal. ±50 ppm or 4 seconds per day
Temperature stability with an expensive crystal. ±10 ppm or about 1 second per day
Temperature stability of the oscillator, ignoring crystal stability. ±5 ppm or about 1/2 second per day
As you can see from this, if the oscillator works, and you get it adjusted reasonably well, other things just don't matter. The easiest oscillators to get working are Pierce oscillators using an HCMOS inverter, or a Colpitts oscillator using a bipolar transistor.
With a 4 MHz crystal, it is quite easy, especially for the Pierce oscillator with an HCMOS inverter if you don't use an unbuffered one, for it to oscillate at 3rd Overtone. If that happens, you need a resistor between the output of the gate and the common point of the crystal and the capacitor.
There have been whole books written about crystal oscillators. Just buy the oscillator.