I realized that most modern textbooks (see Sedra Smith, Razavi, Franco, even AoE) analyze BJTs using parameters such as \$g_m\$, \$ \beta\$, \$r_{\pi}\$, while old books tend to use two-port parameters such as \$h_{ie}\$, \$h_{fe}\$, \$h_{re}\$, \$h_{oe}\$.
So, apparently h-parameters of BJT are no longer popular. Why?
A good question, with answers that will be somewhat opinionated (but here goes, with that caveat)...
Those AC h-parameters are all greatly affected by bias conditions. Presenting those variations in a data sheet is a problem.
Notice that S-parameters are still used widely, especially for MMIC devices run with fixed bias conditions. Measurement apparatus allows S-parameters to be found easily. While H-parameters could be measured (or derived from S-parameters), what would be the point? Those MMIC devices will likely be used in a near-50-ohm system anyway.
Circuit analysis software like SPICE can do a better job than using matrix circuit analysis techniques that would also employ y-z-a-parameters. And SPICE can handle non-linearity too, while matrix analysis cannot.
Perhaps h-parameters were more important when single active devices were costly, and every bit of performance was required of every device. Now we design circuits with ICs having excess gain that is used with feedback to make near-ideal circuits. Somewhat brute-force, but widespread.
In prior decades, 4-param descriptions of active elements was very powerful and was a rite-of-passage for analog designers. Note the 4 hybrid params do not include the various capacitive parasitics, thus any AC performance is questionable; understanding when the various capacitors (e.g emitter diffusion versus emitter drift capacitance) become important is yet another rite-of-passage.
To actually design a circuit, the DC and AC device params, used on back of envelope thinking, are needed.
Otherwise, just let the "optimizer" run overnight.
