Dear Square Enix: Please Remove the Glamour Dresser Cap

[Atelier-Bagur]

Honestly, yes. There are so many issues that could be solved with a codebase overhaul.

The dev team have been very open both in word and design that they don't expect us to be playing this game all the time, so an extended break to fix all this stuff probably wouldn't be too bad, for both us and them. I'm sure they get just as frustrated as we do sometimes with the code limitations.

Likely the Squenix execs/shareholders wouldn't allow it though. No way they'd let their cash cow go for that long.
Suffering from success once more I suppose.

Pretty much this. The only possible reason they would have to be forced to shutdown for an engine overhaul is only if they start to lose both money and subscribers just like with 1.0. But fortunately and I suppose unfortunately in our case where we want this to be fixed, that isn't happening anytime soon considering the general success of FFXIV

[Larirawiel]

Small idea of sorts.

Given that the glamour dresser likely isn't going away, what about an in-game 'catalogue'? One where you could maybe 'sacrifice' items to an NPC, and get a 'mark' in exchange that lets you rebuy that gear (or a replica of it) for say, 1000 gil?

That way, people aren't clinging onto old gear for the sake of glamour, it frees up inventory/retainer space, and also lets them replace items at will if they need them for a specific glamour?

Bear in mind that most people who own additional retainers are likely not doing it for the sake of holding a few extra glams, but because they're omnicrafters who want 10 retainers gathering materials for their Day 1 Savage precrafts, whilst they're petting animals in Island Sanctuary, so I don't think SE would lose money from it ^^

I guess, they cannot separate the appearance of an item from the item itself. This happens when the database is not normalized very well. So if you do not own an item in one or other way then they cannot retrieve the apearance of it. If you want something like that then you have to modify the database structures. And it seems, that they struggle very hard with it. It is possible but it can be very error prone and time consuming.


Cheers

[Atelier-Bagur]

Btw I suppose this talk could be also a reason why Yoshida has made statements about wanting to make another MMO. Probably because he and the devs never got a chance to make one fully from the ground-up.

[greatpumpkinator]

Hmm, not sure how much of that is true, but lets math it out.
A boolean variable generally takes up 2 bytes in an sql database. So lets go with 2 bytes as a true or false if you made the glamour system like WoW's. Where you can collect every look automatically just by equipping it. Then you don't need a seperate storage database you just need one list that matches up to the total item list...so just one boolean, ie a yes or no on the have you collected this look list.

So for say a 400 booleans would just be 800 bytes per user which would work out to 35.8 million subscribers * 800 bytes = 28.64 GB. That is the total size if it is just a boolean for every active account. Now because it is not attached to the total list you might also need a number in there to record the data of the look to reference so lets double it and just say 56GB

Ok so how big if we apply it to a list of every item? Well how many items are there that have looks? I looked on the erozea database and found this.
Arms 4693
Tools 1007
Armor 11215
Accessories 3928
total 20913

So with that we could just take the entire database of those items and add a yes collected look column, or a new table that corresponds to the full look list. So...
35.8 million subscribers * 20913 equipable items *2 bytes = 1497.37 GB for the entire currently active population. It would require account wide sharing of the unlocked looks, but I also don't think every item has a unique looks so I think it could be done. Say double for old accounts, and that is totally doable. Each server would only need an extra TB and they would have a ton of space to handle it.

So storage size isn't it, but technical systems might be I just can't figure out what weird reason that would be though.

A list of boolean values is not the correct data structure, I would argue. If you store the dresser as just lists of boolean values, you're trapping yourself into a non-expandable situation. e.g. In the list of "Head" slot items, Item #1 is ItemA, and Item #2 is ItemB. A mapping of list positions to the item it corresponds to would be needed, and that's not sustainable, or maintainable IMO. At a minimum, you're going to want to store it as the ItemID, which is likely a UInt64. That's 4x the size of the boolean...but still probably not so bad.

And to add to that, you can dye a piece of armor first, and store it in the dresser, and that dye remains. I would hate to lose that, so now you're talking at least adding the DyeID in with that, which is probably another UInt64 value.

[Ransu]

Then they lied. Because they admitted, that the 2002 error issue during the EW launch was caused by some old 1.x code in the client.


Cheers

The game itself likely doesn't have any 1.x code in it. The log on servers are a different thing entirely. The launcher still uses MSHTML, so its not that hard to believe they re-used 1.x code for the log on servers and other networking structures. You all gotta remember that ARR was built in like 2 years. That's extremely short dev time for an MMO, so obviously some code in other areas outside of the game are gonna get re-used. They also obviously see the rush to build ARR is catching up to them which is why we're seeing so much talk about upgrades coming in 7.0.

[Iyrnthota]

The game itself likely doesn't have any 1.x code in it. The log on servers are a different thing entirely. The launcher still uses MSHTML, so its not that hard to believe they re-used 1.x code for the log on servers and other networking structures. You all gotta remember that ARR was built in like 2 years. That's extremely short dev time for an MMO, so obviously some code in other areas outside of the game are gonna get re-used. They also obviously see the rush to build ARR is catching up to them which is why we're seeing so much talk about upgrades coming in 7.0.

They have been upgrading server infrastrucutre recently IIRC, might be they're well on their way to purging some of that old 1.x stuff.

[Larirawiel]

So...
35.8 million subscribers * 20913 equipable items *2 bytes = 1497.37 GB for the entire currently active population. It would require account wide sharing of the unlocked looks, but I also don't think every item has a unique looks so I think it could be done. Say double for old accounts, and that is totally doable. Each server would only need an extra TB and they would have a ton of space to handle it.

So storage size isn't it, but technical systems might be I just can't figure out what weird reason that would be though.

Sorry, but this is not true.

For booleans you need one bit and not two bytes. One byte is 8 bit. So you could store true and false in one byte. Take the first bit of a byte and flip it to 0 or 1 and you are done.

And your assumption, that there is one or two bytes for every item for every player in the game would be very inefficient in terms of storing and retrieving the data. The cardinality of the true/false column would be very low and this would propably lead to so called "full table scans". And full table scans are very slow. Especially when you have this huge amount of data.

And this is the reason why almost nobody would store every item and every player and put a true/false-flag on it. Usually, it is made with a so called "m:n"- or many-to-many-relationship. It is way more efficient to store and retrieve the data. Because you would only store records of items, which the player really got. And the bridge table between player and item would contain many different IDs. The cardinality would be way better compared to a true/false column (yes, i know, some databases have special indexes for booleans) and a database index could be used.


Cheers

[greatpumpkinator]

Sorry, but this is not true.

For booleans you need one bit and not two bytes. One byte is 8 bit. So you could store true and false in one byte. Take the first bit of a byte and flip it to 0 or 1 and you are done.

Cheers

Yes, you _could_ do that (I've even done it myself years and years ago because it was fun to be able to store 8 boolean values in 1 byte). We have no idea of how SE is storing this data, do we? Without knowing that, this is all just speculation. That said, I wouldn't put it past them to do the bit flipping, but storing it as a single byte is probably more efficient (CPU-wise at the cost of storage space).

Edit: Looking at some...uh...online code related to FFXIV that I won't mention...I see bytes used as both a standard byte, and a bitmask, which infers that they are, at least in some respect employing the bit flipping. But that still doesn't mean that they are storing booleans to disk that way. To be honest, using the bit flipping to store true of false on if an item is in the glamour dresser would be a maintenance nightmare. So, 1 byte would store 8 items...and how are you mapping those items to the respective ItemIDs? I posted earlier in this thread as to why using a simple boolean is a bad idea for this.

[IkaraGreydancer]

you should not need to buy more retainers to store weapons and gear.
the glamour dresser should be able to store everything.

I mean..
You shouldn't be using the dresser as a storage period. That's not it's purpose

[Theodric]

I mean..
You shouldn't be using the dresser as a storage period. That's not it's purpose

It's a viable purpose when competitors have a built in system to allow people to change their glamour at will based on what they have acquired thus far in their respective games. People like collecting things, which is precisely why the carrot on a stick for any new content typically includes new glamour amongst other collectable rewards such as mounts, minions and emotes.