w3c / danmaku

Bullet Chatting Proposal

Introduction

This repository contains the use cases and proposed API for Bullet Chatting.

• Bullet Chatting Use Cases (中文版)
• Bullet Chatting API Proposal (中文版)

Demos

We also have several demos:

• simple-demo
• bulletchatting-proposal-demo
• Demo without video

Misc

Other documents are in the bullet-chatting-docs repo.

danmaku's Issues

About the "Deterministic of rendering"

Current specification describe that "the position and order of each Bullet Chatting is fixed each time it is rendered". When "allowOverlap" is set to false. The position of each bullet chatting is relay on the previous chatting's. User agent have to calculate the position of all chatting's from the very beginning of the list for a certain chatting, which may cause performance issue. And if client try to insert a chatting before current timestamp, should the user agent repaint all chatting's currently shown to fit this change? As I know, There is not any website rendering chatting's like this currently. So, why this characteristics is listed?

Use cases that do not involve web applications running in a UA

(From https://lists.w3.org/Archives/Public/public-web-and-tv/2020Jan/0000.html)

Do we have use cases that do not involve web application running in a user agent? If so, we can probably mention them in the use cases document.

Rendering determinism and ability to pause a comment

I'm still wondering how a possible "pause" mechanism affects rendering of a bullet chatting experience.

According to the spec, one of the characteristics of a bullet chattting experience is "Deterministic of rendering", but if an application supports a "pause" feature for an individual bullet chatting comment, wouldn't the rendering determinism be affected when a user pauses a comment?

In the spec, it may just be a matter of completing the definition, for instance:

<li><dfn>Rendering determinism</dfn>: Provided that the bullet chatting
<a>container</a> and the <a>bullet chatting comment</a> are fixed, and
in the absence of user interaction (e.g. to pause a <a>bullet chatting
comment</a>), then the rendering position, order and timing of <a>bullet
chatting comment</a> are always the same.</li>

(I also suggest to rename the characteristic as "Rendering determinism" which sounds more English to me)

把弹幕变成规范这个提案的目的是什么？

Issue by LongTengDao
Jul 18, 2019, 8:33 PM GMT+8
Originally opened as w3c-proposal-incubation/w3c-proposal-incubation.github.io#4

原生化，以提升性能？
还是只是某种具体实现（或许是最佳实践）的公开文档？

Accessibility of danmaku (bullet chat)

From https://lists.w3.org/Archives/Public/public-web-and-tv/2020Jan/0007.html :

Accessibility is always an important consideration for any W3C specification work. Have you considered how to make the Bullet Chat comments accessible? This would be an issue with canvas based rendering, for example.

What does "survival time" mean?

In Uniformity of modes, what is the "survival time" of a mode?

Do you mean to say the duration of a bullet chatting comment?

Or is this talking about the lifetime of the set of layers that have the same mode? If so, is it different from the sum of durations of all bullet chatting comments?

Time synchronisation between live stream and bullet chatting

From https://lists.w3.org/Archives/Public/public-web-and-tv/2020Jan/0007.html :

In a typical bullet chatting implementation for live streams, are comments distributed to clients in real time, e.g., over a WebSocket connection, and displayed immediately when received, without being synchronized to the media timeline?

As each client may be at a different playback position on the media timeline, is any timing information used to ensure the comments are relevant to each viewer? For example, what happens if I'm watching the live stream but 5 minutes behind the most recent playback position?

What does "timeline" mean?

I fail to parse the definition of the timeline term. Currently, it is "The on-demand barcorder is a real-time insertion or custom timeline for video playback time, live broadcasts, and other scenarios".

For instance, "barcorder" does not mean anything to me, and if it's a typo for "barcoder", I don't understand what a "barcoder" has to do with bullet chatting.

Do you mean to say that there are two types of bullet chatting comments, those that get rendered as soon as possible without any reference to the media timeline, and those that start at a specific point on the media timeline?

Also, what is the relationship/difference between this and the concepts of "appearance time" and "duration" in Basic properties?

Independence of space and layers

The Independence of space characteristic describes a system of layers that does not seem to appear anywhere else, but that seems actually important to understand how things get rendered in practice.

Are the following assertions all true?

1. A bullet chatting experience is composed of a set of superposed layers.
2. A bullet chatting comment belongs to one and only one layer.
3. A layer only contains bullet chatting comments that have a given mode.
4. Within a layer, bullet chatting comments never overlap (unless a user pauses a comment?)

In particular, is 4 true? I'm asking because the proposed API includes an allowOverlap attribute. Or should 4. be more "does not overlap unless there are too many comments to render and overlap is allowed"?

If 1-3 are true, I believe it would be useful to introduce the term "layer" more formally in the spec.

Attributes vs. Styles

play-state, duration, delay are three styles listed in current draft. Why they are styles instead of attributes?

play-state looks like HTMLMediaElement.paused to me, which would be an attribute.

It is normal that delay of each comments are different, and using style for it is confusing.

『滚动弹幕』和『逆向弹幕』的命名不对称

感觉改为『滚动弹幕』和『逆向滚动弹幕』会更清晰一点。（实际上正文里已经有地方使用的是『逆向滚动弹幕』的写法，建议统一。）

对应的，mode属性reverse建议改为scroll-reverse。

请留意在日本地区的专利保护问题

众所周知所谓的弹幕起源于 niconico 动画 “自有” 的评论功能。其实关于这种评论的 “显示装置、评论显示方法以及程序”（表示装置、コメント表示方法、及びプログラム）早在 niconico 动画创立伊始就由其运营方株式会社 Dwango 向日本国特许厅申请了发明专利（于 2006 年 12 月 11 日申请，公开特许公报于 2008 年 6 月 26 日公开）。日本特许法第六十七条规定其专利（特许）有效期为自申请之日起二十年，并可择期延长。W3C 作为互联网国际标准的制定者，本规范草稿存在的专利隐患可能是亟需讨论的。在正式发布本规范之前请三思。

非相关专业出身，若有措辞不当请见谅。

allowOverlap attribute for comments in different modes

As the draft described, there are 4 modes of bulletchat element. The example only shows how to handle scroll mode bulletchat elements when allowOverlap is set. May a bulletchat[mode="top"] overlap bulletchat[mode="scroll"] when allowOverlap is set to true?

Link to English explainer on Danmaku

It would be helpful for people who are unfamiliar with Danmaku to make an informative reference to a page which explains this; preferably in English.

I found this page helpful: http://danmaku.weebly.com/ but feel free to link to a better one.

[uc] use case proposal: Bullet chats for 360 videos

As I mentioned several times, e.g., during the MEIG F2F meeting at TPAC, I've been wondering about a possible furure use case for bullet chatting like putting bullet chats on 360 videos and specifying the position/geolocation/orientation of the bullet chats on the 360 screen using WebVMT, etc.

Scope of the task force

https://github.com/w3c/danmaku/blob/master/TF/Bullet_Chatting_TF_Draft.md

Currently the scope is "TBD". We should make it clear.

Any API to seek bullet comments list?

It is normal that bullet comments are used on a <video> element. And user may seek the video to some timestamp. This require an API to seek the bullet comments too. Is there any API for this action?

scroll direction should respect language script direction

value scroll and reverse should respect language script direction. As in most ltr languages it should scroll as described in current specification. But it should be reversed in rtl languages. CSS direction should be used in this context.

Discuss: What should the render do if comments use languages with different script directions?