[MENTION=4937]What you do not take into account is that the Longbow is totally ineffective versus any sort of plate armor, mostly ineffective vs. chain and even with padded armor you would stand good chances to be unhurt if the longbow hit you on a covered body part.
Wait, what? Longbow points go right through anything but double layer mail ('chain') without hardly slowing down. The rise of the longbow and the crossbow was a big reason behind the rapid development of plate armor. The longbow will absolutely penetrate a gambeson ('padded') without much difficulty. While a gambeson will dissipate a lot of the energy of the shaft, you're still going to end up with a barbed arrow sticking in you 4-5" deep. Remember, many 13th and 14th century knights would be wearing mail over a gambeson, the longbow would frequently penetrate the combination to a depth of 2-3" (driving cloth and broken rings into your body as well), especially at ranges under 75 yards. And given the rate of fire, once you started to get hit and were now in shock and debilitated, you'd quickly turn into a pin cushion.
Plate was designed to resist this, but it was itself also penetrable, particularly plate of the quality available to all but the most wealthy which would have gaps, thin flat areas, and lower quality steel. Coat of Plates ('Brigandine' in D&D parlance), which modern D&D doesn't really have, largely replaced mail coats in the infantry precisely because they would stop arrows from penetrating reliably and could be constructed cheaply. Plate over gambeson would likely stop arrows from penetrating, but needle point arrows designed for penetration still can punch through plate at close range - albeit likely without enough force left to seriously hamper their target and not consistently - an occasional scratch as you put it.
D&D hasn't attempted to model this at all since 'weapon vs. AC modifiers' dropped out of favor, which incidentally (moving back to some earlier comments) did attempt to model the relative futility of using an arming sword (D&D's inaccurately named 'longsword') against plate armor.
As for the primitive fire arm, you might be surprised that for all but the heaviest firearms available, the results would have been similar. Wheellock muskatoons and pistols had muzzle energy in foot pounds not that much higher than longbows (the pistol) and windlass operated crossbows (the musketoon), and plate armor of the period was designed to deflect these devices. Indeed, plate was 'proofed' by firing a pistol at the breast from point black range. The resulting dent proved to the customer that it could deflect a bullet.
Remember back then injury meant high risc of infection and people dying from arrows which according to some historians was the highest fatality cause in warfare back then...
In WW1 as well, for that matter. Almost everyone that is injured dies from one of three things, none of which is modelled by D&D (or any other commonly played RPG) - shock, blood loss and infection. Until your are modelling blood loss, either from obvious wounds or internal bleeding, and modelling infection you aren't really trying to be realistic. And RPGs, even those with pretensions of realism, really aren't trying to be realistic, because it generally doesn't make for a fun game.
All of these lead to compromises somewhere if you try to integrate primitive firearms. Either you need at least give the m the highly unrealistic RoF of crossbows or some other gimmick. Would you just up the damage say they do 4d10 but need 4 rounds to reload, Then they will only be used as a opener for a first round of combat, because sitting out 4 rounds in 5e means the combat might be over already. So if you want to introduce them in a meaningful way, at least make them stand out in terms of usefulness versus other ranged weapons.
e.g. give them an advantage for the ini roll make them ignore armour, exploding dice etc.
This decision, nominally motivated by realism, results in an ahistorical reality where the firearm obsoletes other weaponry in short order, rather than being one component of a combined arms force as it actually was right up until the late 19th century (though it would take many years before tactics on the battlefield would catch up to this fact). Hand gonners, musketeers, and fusiliers did not drive other sorts of troops from the battlefield immediately. The weapons those troops employed had many disadvantages. Typically what you saw developed was an assault force wielding firearms, protected by a screen of pikemen or other troops that could provide protection while the firearm wielders clumsily reloaded their slow firing and frequently inaccurate weapons.
If you do not care for this then the following might happen in your game: Let us take a standard situation in many movies and put that on a RP base: One guy threats the other with a loaded gun
It would be ridiculous to do this in 5e when the gun only did 1d8 on a standard attack Rolland then need some round or two to reload. Every level 2 and up PC would just ignore this threat.
Yes, but this example is ridiculous. While what you say is true, the same problem occurs with trying to model a movie scene where a man holds a knife to the other mans throat. A knife slitting your throat will kill you just as quickly and maybe more quickly than a gun shot wound, yet if you follow the RAW strictly this attack only does 1d4 damage and isn't necessarily lethal versus a peasant damsel.
The problem with your example isn't how the gun is modelled, but how damage generally is modelled. D&D's ablative hit points only conceptually work as a model if fortune is in the middle of the action/resolution cycle. The problem with things like falling, exposure to lava, and someone getting a drop on someone with a weapon is that we've removed ourselves from that model, and instead put fortune at the end. In other words, we have already set forth what happened and now we are trying to model it. But D&D's hit point model depends on modeling first and then only afterwards setting forth what happened. It simply can't deal with a process where what happened is known and we try to then simulate the results. Thus, D&D has always applied something like a 'coup de grace' exception to the rules, where if the outcome is actually known, the normal rules of damage simply don't apply - the character is simply dead. However, in the interest of not making this routine, the game typically does not apply many exceptions and makes achieving the conditions of that exception difficult.