RTL versus DTL
Both RTL (Resistor Transistor Logic) and DTL (Diode Transistor Logic) were common logic families in the early days of transistors. The image above shows inverters in both logic families.
RTL is the simpler of the two, but it suffers from many issues: The L->H threshold is defined by the forward voltage of the base-emitter junction of the transistor and is hence rather low, around 600mV for a typical silicon transistor. Furthermore, if the input is at high level it will sink current which is supplied by the collector resistor of the preceding gate. To maintain consistent logic levels it is therefore necessary to adjust all resistors according to fan-out. Besides that, there is also a lot of nastiness in dynamic operation.
DTL addresses these shortcoming by introducing additional diodes: When the input is high, all the base current will be supplied by the base resistor (R5) and there is no current flowing into the gate. When the input is low, current is sourced from the gate, which is sunk into the transistor of the preceding gate. Since the transistor is fairly low-ohmic when it is turned on, there is much less limitation in fan-out. Furthermore, it is not necessary to adjust gate resistors depending on fan-out. This allows using the same resistor values throughout the entire circuit.
A second improvment in DTL is to lift the L->H threshold to a higher voltage. It is now defined by Vbe+Vd2+Vd1-Vd3, which is roughly 1.3V depending on the compenents used.
From DTL to LTL
The image above shows the circuit of an LTL inverter. At the first glance it looks very straightforward: We replace the two silicon diodes with a LED. Since the LED has a much higher forward voltage (depending on LED type), only a single diode is sufficient. This is nice, because it saves one component, but is it actually a good idea?