I think you and Kuroshima are both right. No, I'm not being wishy-washy.

Simultaneous action is something I've been playing with. Setting aside practical considerations ("does it play well?"), simultaneous actions (or declaration of actions) is a perfectly sensible thing to allow in a combat sim, as it of course can happen in real life. In fact, that's a (minor!) failing of typical RPG combat sims, in that the rules simply don't allow for two combatants (two crazed or dumb combatants, perhaps : ) to go at each other simultaneously, and run each other through at the same instant.

BUT, there are some special ingredients needed to make such a thing work. One, likely, is the ability for a character to perform two or more actions simultaneously, such as twisting aside (i.e., dodging) while attacking. Two damn-the-torpedoes swordsmen could each try to skewer the other guy while avoiding his stab – doubtless a difficult thing to do, and very dangerous for each, but it could be attempted.

Another is the addition of an "abort" action, so that when two hotheads jump at each other at once, one could lose the "game of chicken" and revert to panicked defense instead. Again, it'd no doubt make for a difficult defense, but might be smarter than the above crazed move.

And despite cool capability for simultaneous shish-kebabing under the rules, a sword battle still needs to allow for plain old back-and-forth thrust-and-parry. That doesn't necessarily mean a return to ABAB turns; simultaneous action could still work, but again, with the right ingredients. It may require tracking strikes and re-readying as two separate parts of any melee attack, like GURPS does for halberds (but not swords).

Picture it under simultaneous actions. Fighter A makes a thrust, while B (more or less) simultaneously parries it. Done. Then we simultaneously declare actions again, but this time, A can't thrust, he's still "extended". He needs to pull back, while B is free to take some sort of swipe (we assume his parry left him in reasonable position to do so). So A will have to parry that while withdrawing his extended arm (as above, essentially two actions at once: pull back arm + parry).

On to the next actions. After his parry, A is presumably able to launch an attack from his position. B is "extended" and can't attack; he has to withdraw while parrying. And so on, back and forth, clang-clash-clang in classic swordfight style.

All very interesting stuff – as a thought experiment that I haven't seen yet brought to live play. Whether it could work, or whether the minutiae and bookkeeping would kill a real game, are big questions, and that's where Kuroshima's concerns may win out. (Or similar concerns; I don't see any problem with "round boundary", just overall futziness.)

As always, that was longer than I expected. : ) I'll try to expand/expound a bit more in the upcoming second part of the article.