Game:

Guilty Gear -STRIVE- (2021)

Duration: 23:52

27 views

2

Here's where being a test subject doesn't actually pay off!
twitter.com/yuugijoou
discord.gg/bKT9pRW

---

Of course, it's not all smiles and sunshine in the kingdom of Parsec. There are indeed cases where it's vastly preferable to do the thing the game is designed to do, rather than come up with overly convoluted and clever means of circumventing exactly that.

The beauty and ultimate purpose of Parsec is NOT actually trying to avoid having to use less than ideal networking processes implemented in games that actually DO have online play. The fact that it actually DOES allow for this in fringe cases IS patently fascinating, at least, but it's more of a byproduct... and it's also always going to be far from ideal.

The hosting player will always be at an intrinsic advantage, for one. I'm actually playing the game as normal on my own machine. So there's no technical struggle to translate my actions to the screen whatsoever. The other players, however, will have to contend with at least some theoretically significant issues. At its most basic, they ARE going to be suffering input delay, one way or another. They DO have to send me their game instructions so that MY game session can know that they've acted. They also have to wait for video feed of what they're looking at to reach them... and, it being a video feed, after all... it too will be subject to network travel time.

The reason why this is TECHNICALLY not the worst idea in the world insofar as existing networking solutions go is that Parsec is designed with the utmost efficiency in mind. Sometimes at the expense of all other concerns. The video feed may suffer for the connected client player(s), and they all have to share the collective bandwidth allocated by the host, but Parsec is designed to avoid stressing out over dropped frames and strange artifacts... normally a cardinal sin in video and video GAMES, all for the sake of the smoothness of experience.

Video games themselves as products of a lucratively large and discriminating industry are not designed with this in mind, because as packaged and sold experiences, they'd be laughed off the market if they conceded their primary gameplay considerations for the sake of such things... unless they hide it in VERY clever ways. Which a lot of the cleverest DO! And even THOSE catch a wary and watchful eye, curiously bemused at the existence of such things in the first place. Human beings are, by nature, rapid natural processors of sensory information with highly discriminatory reflexes they're often not even fully aware of... they job of anyone working in audiovisual presentation is to learn how to fool them most effectively. Film and television have been doing this for actual generations now!

Furthermore, your netcode implementation is trying to tackle a far more nebulous networking task... you're actually trying to take TWO games and somehow manage the feat of synchronizing their execution and logic such that it appears as though two people are sharing the same game, rather than nudging the pieces along in each other's in response to the opposing player's actions. It's not through sheer lassitude alone that things have stagnated into "well, it works, and it's not AS awful as everything else we've tried to get this far!" ...ANY netcode alone, let alone functionally acceptable netcode is a downright herculean task.

That's what makes rollback netcode so special. It's an entire paradigm shift that most implementations simply aren't ready for. That needs to be put into place very particularly and thoughtfully from the foundation up. It demonstrably CAN be retrofitted into an existing implementation, but it's done so at a VERY labor-intensive fashion that more or less requires a behind-the-scenes overhaul of how the game operates.

And of the two flavors of sorcery, it's no wonder that rollback IS the superior choice. It's NOT beaming entire video streams to someone else, shoving the network burden off of one player at the direct expense of any others. Solutions like Parsec can, by their very nature, never actually compete with that... on a theoretical level. On a purely theoretical level, indeed, every client is POTENTIALLY subject to the same awful deal as delay-based netcode has apparently found or forced to accept as tolerable. And that worst-case scenario is enough to put many off the prospect entirely.

However, in practice, I'd say that, at least in my experience so far, as both a host and a client... the truth is somewhere in between. All those visual concessions (which aren't going to be frequent, assuming the configuration is sound and if the network is stable... which NO netcode model can guarantee, mind you) mean that at the absolute WORST case, you'll have some hiccups in your controls being sent back to the main game... or perhaps some visual muddling that will make it potentially more difficult to play... but the remote host paradigm won't make you actually WAIT to play by force.