wow Dev Watercooler: WoW Classic Development

[Stan 5,240]

Blizzard released a new Dev Watercooler today with information about the ongoing WoW: Classic development. Classic servers will run on Patch 1.12: Drums of War, because it represents the most complete version of the classic experience. Devs are hard at work to make the modern client compatible with the classic data.

Blizzard (Source)

Greetings! Development of World of Warcraft Classic is underway, and we’re very excited to share some of the challenges and solutions we’re working on. As we mentioned last BlizzCon, the process of restoring the classic game is not straightforward, and it’s important to us to take the time and effort to get it right—this includes poring over numerous game versions, data, and code; meticulously scrutinizing all the changes we’ve made over the years. Rest assured: The WoW Classic team is hard at work making it a reality, and we’re at a point in development where we’re ready to share some of the things we’ve been working on.

WOW CLASSIC: FIRST PROTOTYPE

The first—and among the most important—decision we had to make was which version of the game to focus on. As many of you have noted, the classic period was two years long and full of changes. Core features like Battlegrounds were introduced in patches after WoW’s original launch, and class design similarly changed over time. After careful consideration, we decided on Patch 1.12: Drums of War as our foundation, because it represents the most complete version of the classic experience.

Once we had our starting point, we began taking stock of what we had in the source code and what we could make available, which included restoring the original development database from archival backups. After stitching various key pieces together, we had a locally rebuilt version of Patch 1.12 running internally. The team could create characters and do basic questing and leveling—and dying, which we did many times. For testing purposes. Obviously.

Our initial runs exposed a few (expected) issues: the game sometimes crashed, didn’t recognize our modern video cards, and was incompatible with our current login system. That first pass also couldn’t support any of our modern security and anti-cheating capabilities. Clearly we had a lot of work to do to make WoW Classic live up to the Blizzard standard of quality, and deliver the experience players want.

THE PATH FORWARD: SECOND PROTOTYPE

Speaking of engineering, World of Warcraft is a very data-driven game, which means the basic code is flexible and the specific way it behaves is controlled by information contained in databases. Things like quests, monsters, items, and the rules for how these all interact are defined by the designers and artists in data.

So we asked ourselves, would it still be possible to deliver an authentic classic experience if we took our modern code, with all its back-end improvements and changes, and used it to process the Patch 1.12 game data? While that might seem counterintuitive, this would inherently include classic systems like skill ranks, old quests and terrain, talents, and so on, while later features like Transmog and Achievements would effectively not exist because they were entirely absent from the data. After weeks of R&D, experimentation, and prototyping, we were confident we could deliver the classic WoW content and gameplay without sacrificing the literally millions of hours put in to back-end development over the past 13 years.

While our initial effort helped us determine the experience we wanted to provide, this second prototype really defined how we’d get there. Starting from a modern architecture—with all its security and stability changes—means the team’s efforts can be focused on pursuing an authentic classic experience. Any differences in behavior between our development builds and the patch 1.12 reference can be systematically cataloged and corrected, while still operating from a foundation that’s stable and secure.

DIGGING IN

So what does it take to recreate an authentic classic experience with modern engineering? Let’s start by categorizing the different types of game data that make up WoW:

• Table data: This kind of information is almost always represented as numbers. How many hit points a creature has, the amount of Strength an item grants, or where and when certain creatures spawn, are all examples of the numerical data we store in our databases. We can also store and enforce relationships between different pieces of data.
• File data: This is often very dense data like 3D models, textures, animations and terrain. Our user interface is built up from XML and Lua files. Many of the art files do not use the same file formats that commercial art tools spit out. Our build pipeline takes these raw art files and translates them into something optimized for our game to read and process.
• Lua scripts: Some features are driven by Lua scripts written by designers, allowing them to easily define custom behaviors for server-side logic without requiring deep engineering knowledge.

HOW ENGINEERING HAS CHANGED

One challenge we face is that all the classic data is in the original format used at launch, but that format has changed substantially in the intervening years. Major work needs to be done in this area to make the modern client compatible with the classic data.

For example, spells could originally only perform three actions on the spell’s target. In table form, that looked something like this:

ID	Name	Effect One	Effect Two	Effect Three	Aura One	Aura Two	Effect Damage One	Aura Damage One	Aura Damage Two
1	Fireball	Deal Damage	Apply Aura	Nothing	Nothing	Deal Damage Periodically	30	Nothing	3
2	Frost Bolt	Deal Damage	Apply Aura	Nothing	Nothing	Slow	20	Nothing	Nothing

As you can see, there is a lot of space taken up by ‘Nothing’. Over the course of WoW’s lifetime, we’ve improved our data design and normalized much of our database data. Today, that same data would be separated out like this:

Table Name: Spell

ID	Name
1	Fireball
2	Frostbolt

Table Name: Spell Effect

ID	SpellID	Effect	Damage
1	1	Damage	30
2	2	Damage	25

Table Name: Spell Aura

ID	SpellID	Aura	Damage
1	1	Deal Damage Periodically	3
2	2	Slow	Nothing

In this form, there is much less wasted space and spells are no longer limited to three effects. But before we can load any database data, we need to transform the old data layout into the new one. This is not limited to spells, as almost every game system (including items, creatures, player characters, spawning, AI, and more) has had its database layout altered over the years.

LOOKING AHEAD

All the work we’re doing will ultimately allow us to recreate an authentic classic experience on a platform that is much more optimized and stable, helping us avoid latency and stability issues. Additional modern improvements will include modern anti-cheat/botting detection, customer service and Battle.net integration, and similar conveniences that do not affect the core gameplay experience.

We are looking forward to the challenges ahead and share your passion for the classic game; every code check-in data conversion we make brings WoW Classic closer to providing that authentic experience you—and we—want. Thanks for joining us on this journey.

[Laragon 269]

Huge fan of this information. It makes me feel like the time line is even further away though.  Still cool to see behind the scenes things they've learned and things they are yet to discover. Never even realized how their coding of spells has changed so drastically, even on things like frostbolt or fireball that are original spells.

[jinsu2301 273]

This is exactly what I was hoping they would be doing: take 8.0 and turn it into 1.12. Curious as to whether or not they will use sharding and maybe even a single mega-server since it would help massively with fluctuating server populations. 

[Abom 31]

This information is the difference between console and PC developers; Gettin' all gangster with the engineering.

I just want my Windfury Totem back, daddy

Edited June 15, 2018 by Abom

[tkioz 229]

This is great information and really shows the difference between producing a stable professional product and hacking together a kludge that private servers are.

[thurk 52]

I can't believe it's taking so long, if all they had to do is change frostbolt and fireball.  I can see it's a complex change, but maybe just release WoW Classic without mages for now?

(omfg I'm just kidding, please stop hitting me.)

[ionix 8]

It was obvious that they can't just boot the 1.12 server from backup and then implement it. Too many legacy code and bugs.

And supporting two different code bases really sucks. So, what they came up with is the obvious solution - just use the base of the game, but scale it to 1.12 content. This way they can implement future stability/backend patches to the Vanilla server as well, greatly reducing the time and effort required to support it.

However, while just booting the 1.12 server may have taken like a week, the current project is huge. I will not be surprised if it takes them 2 years or more. It's just a ridiculous amount of work.

It gives me hope that we may see TBC and Wrath though. While Vanilla could take quite a bit of work to roll back due to the Cata rework, TBC and Wrath are pretty much unchanged in their current version on live.

Edited June 16, 2018 by ionix

[Seksi 208]

Think that having "past WoW" at our fingertips becomes much more possible once Classic (and its lessons) are finished. Looking forward to a lot of nostalgia (and new WoW content!) for the next few years :)

[Arcling 898]

On 6/16/2018 at 11:02 AM, ionix said:

It gives me hope that we may see TBC and Wrath though. While Vanilla could take quite a bit of work to roll back due to the Cata rework, TBC and Wrath are pretty much unchanged in their current version on live.

Mechanically they are, several bosses, other than tuning the numbers, had changes to become soloable. Rolling it back wouldn't be that easy. I guess like Vanilla, it might require some reverse engineering of sorts, since game is running on a different structure. Anyway, that's unlikely, as it might split playerbase way too much, and it might have ended with dead servers.