Eldarth wrote: »

Congratulations. Animation Csnceling, macroing, and action weaving cause lag.

You've at least tripled or note the network traffic to and from the server to every player in the area.

Got lag? ...how surprising.

Interesting. Any proof of concept? Or any proof at all?

How does macro-ing works? Is macro-ing simply automating animation cancelling?

What's action weaving?

Obscure wrote: »

ZOS loves to embrace their design flaws:

- Block casting? FEATURE
- Animation Cancelling that doesn't also cancel the action? FEATURE
- Infinite horse sprinting? FEATURE
- Cyrodiil bonuses from a campaign you never play in? FEATURE
- Zoning to Cyrodiil from anywhere to avoid paying to wayshrine directly? FEATURE
- Suicide rushing a resource camp the fastest method to travel in Cyrodiil? FEATURE

One man's buggy design oversight, is another man's FEATURE.

Disgruntled player base? FEATURE

Eldarth wrote: »

Congratulations. Animation Csnceling, macroing, and action weaving cause lag.

You've at least tripled or note the network traffic to and from the server to every player in the area.

Got lag? ...how surprising.

Speaking of lag......

So Cyrodil hardly has any deer and from what I've heard they had to remove them and torch bugs to help prevent lag there. If that's true, then why the H#LL did ZOS put frigging Pets in the Crown Store. Not only have you just added another lag causing item to the game people have their pets out constantly. So with a Sorc, they have 3 followers now instead of 2. Just how exactly is that helping?????

Robbmrp wrote: »

Eldarth wrote: »

Congratulations. Animation Csnceling, macroing, and action weaving cause lag.

You've at least tripled or note the network traffic to and from the server to every player in the area.

Got lag? ...how surprising.

Speaking of lag......

So Cyrodil hardly has any deer and from what I've heard they had to remove them and torch bugs to help prevent lag there. If that's true, then why the H#LL did ZOS put frigging Pets in the Crown Store. Not only have you just added another lag causing item to the game people have their pets out constantly. So with a Sorc, they have 3 followers now instead of 2. Just how exactly is that helping?????

You can't have cosmetic pets in Cyrodiil...

Davadin wrote: »

Interesting. Any proof of concept? Or any proof at all?
How does macro-ing works? Is macro-ing simply automating animation cancelling?
What's action weaving?

40 years software development experience.
35 years communication and networking.
35 years low-level hardware, I/O expert.
34 years Arpanet/Internet experience
30 years client/server and three-tier expert
25 years software architect
Professional developer in 10+ different computer languages
Member of Mensa
Will that do?

It's pretty obvious - for animation canceling, the client computer has to send 1)initial action/target, 2) Canceled/New action/target, 3) animation cancellation all affected clients, 4) damage update for (1), 5) damage update for (2).

That's at least 3-5 network interactions between ALL affected clients and servers in the space that should (without animation cancelling) take 1-2 network interactions. Pretty obvious you've at the very least doubled your network overhead for every player within the combat range.

Macro-ing is setting up a single key or mouse button to issue the initial action, perhaps followed by a few microseconds of delay followed by the animation canceling attack action.

Action weaving is setting up macro sequences to maximize animation-canceling actions thereby increasing DPS.

Animation canceling is an exploit regardless of ZOS "acceptance" of it.

Imagine you're in a boxing ring and you know your opponent slight dips his left shoulder whenever he throws a right punch. Now, imagine you're in that same ring and you don't see your opponent move at all -- you simply spontaneously get a bloody nose and a headache.

Or perhaps I move my hand a tiny bit -- and suddenly you're hit by a critical fire spell.

If there is no animation to telegraph what combat action or spell someone is casting then there is no way you can cast an interrupt or block what you can't see.


I'm pretty confident that animation canceling is a major contributor to lag in Cyrodil. If you want proof - gather a couple guilds together on opposite sides and try having a battle without anyone doing any animation cancelling. I'll bet the lag is significatnly reduced.

Prove me wrong @ZOS_Eric, @ZOS_ChrisStrasz

Triddle wrote: »

That's really good proof there. The popular proof technique 'appeal irrelevantly to perceived authority' is employed perfectly.

Kind of hard to prove something from this side of the wall.
If you'd like to decipher their undoubtedly encrypted network protocol packets so we can gather the actual statistics without being employed by ZOS, let me know when you have them and I'll analyze the networking traffic.

Or, if you can provide two guilds on opposite sides to "prove" it by not using animation canceling let me know that too.

Otherwise, I can only rely on decades of experience.
Which is also why I challenged ZOS developers to prove the assertion incorrect.

Eldarth wrote: »

Triddle wrote: »

That's really good proof there. The popular proof technique 'appeal irrelevantly to perceived authority' is employed perfectly.

Kind of hard to prove something from this side of the wall.
If you'd like to decipher their undoubtedly encrypted network protocol packets so we can gather the actual statistics without being employed by ZOS, let me know when you have them and I'll analyze the networking traffic.

Or, if you can provide two guilds on opposite sides to "prove" it by not using animation canceling let me know that too.

Otherwise, I can only rely on decades of experience.
Which is also why I challenged ZOS developers to prove the assertion incorrect.

Whether the packets are encrypted or not does not matter (and of course they are); as per your assertion , the number of packets would increase by using this canceling technique. So, to acquire empirical data you would only need to go on with some of your friends, with a sniffer running on each PC then test with and without. Then *you* can provide us and ZOS the data proving your theory without cracking their crypto package.

Eldarth wrote: »

Kind of hard to prove something from this side of the wall.
If you'd like to decipher their undoubtedly encrypted network protocol packets so we can gather the actual statistics without being employed by ZOS, let me know when you have them and I'll analyze the networking traffic.

Or, if you can provide two guilds on opposite sides to "prove" it by not using animation canceling let me know that too.

Otherwise, I can only rely on decades of experience.
Which is also why I challenged ZOS developers to prove the assertion incorrect.

You did employ a rather interesting strategy there: I can't prove that I'm right, so it's your task to prove me wrong. That is exactly the way to engage in a constructive discussion.

Fun fact: the forums say that the lag started somewhere around 1.2.3 while all those things you mentioned were there before 1.2.3. So, I guess the burden of proof really lies on you.

Don't get me wrong: I would be happy to hear that the lag issue had a clear cause that could at least conceptually be isolated. Yet, I doubt that there is. Recently, I've read about a shift from client-side calculations to server-side calculations that came with 1.2.3 due to heavy cheating. Honestly, that does make much more sense and, additionally, fits well with the chronology.

Zsymon wrote: »

So they need to upgrade their servers for the lag to stop? Or move calculations back to clients so people can play without lag, albeit with cheaters, a difficult decision. To be honest I rather have no lag and a few cheaters, but I guess both problems make pvp pretty much unplayable.

I bet investing in stronger servers would be worth it, since much more people would play then.

If its a cluster they are throttled by the local network/speed capacity...more than the servers.
Then you have to worry about packet load management too.
It would be more appropriate to remove any bottle necks in the cluster operation.

Is it better to have a rapid server with no parallel network (no network latency/protocol issues/orders of magnitude quicker)
Is it better to have multiple slower servers running in parallel (spread the load and add redundancy and distribute clients).

They probably have a good balance between server power and cluster size to maximise efficiency.
So its probably a saturation issue that's is network hardware limited anyway.
Hence "better servers wont help" comment

Give them the benefit of the doubt the cluster is singing.
The only option is to drop packets or reduce packet size (damage data).
Does packet size increase in a large scale battle with lots of AoE ?
I'll get the wireshark out and have a play at some point.
That would make it a transient network load problem waiting for incoming packets

Are those packets bottlenecked in the server/ the cluster / the client network / the client

Eldarth wrote: »

Davadin wrote: »

Interesting. Any proof of concept? Or any proof at all?
How does macro-ing works? Is macro-ing simply automating animation cancelling?
What's action weaving?

<lots of speculation about software the poster clearly can have no knowledge of>

I'm pretty confident that animation canceling is a major contributor to lag in Cyrodil. If you want proof - gather a couple guilds together on opposite sides and try having a battle without anyone doing any animation cancelling. I'll bet the lag is significatnly reduced.

Prove me wrong @ZOS_Eric, @ZOS_ChrisStrasz

TL;DR - I've umpteen years of alleged software development, I have no idea of ZOS' server design and so "I'm pretty confident I know what I'm talking about".

OK.