Meeting minutes
Group activitiess since our last TPAC
AnQing: let's get started
… will not say much about what this group is doing, since every one is quite familiar with this group
… welcome all, and first, let's talk about Review the Chinese Web IG's activities since last TPAC
… give hand it over to Xueyuan
Slideset: https://github.com/w3c/chinese-ig/blob/main/TPAC2025/slides/chinese-web-ig-tpac2025.pdf
Xueyuan: briefly about Chinese IG, the main objective of this group includes identifying requirements and use cases
… build a community of Chiense members to be connected to W3C global community
… didscuss tech dieas and explore potential proposals
… promote standards testing and implementaiton
… organize local events to identify member needs
… this group is in collabration with Chinese Layout TF, MiniApps WG and MiniApps CG
… High-perf Baseline CG
… TAG, WebApp WG, Web Platfrom Incubator CG, Media and Entertainment IG
… Web of Things IG & WG
… also the APA WG
… about activities of this group
… this group does not travel a lot, but it has quite some events in China
… we had a WebEvolve online event in Jan to talk about SOLID
… the WebEvolve event is a W3C event open to public
… about 50 attendees joint this event online
… there is an event report on the website
… in May 2024, this group had an event discussing ML, WebGPU and media related tech
… 20 talks, 90+ attendees on line
… 15000 ppl watch this event on line
… all talked recored and now with subtitles ready
… in June 2025, we had an online event focusing on immersive tech used in internatinal games such as Olympic games and world cup
… China Mobile co-organized this event with W3C
… we also had an event talking about identity tech, standards and policy in CHina
… PLH joint this event
… and the identity community in CHina
… in Sep 2024, Simone joint the Inclusion coference in Shanghai, introducing W3C work on identity and related security work
… there was also an event on DID in late Oct in Beijing, this event called GOSIM
… Drummond the edior of DID WG joint the event, together with other 30 attendees
… Also a NGI track during GOSIM too
… W3C also attended the World Internet COnference in Wuzhen
… W3C had a booth, spoke at the Web tech forum and did out reach to CHnese members who were there
… in Jan 2025, we had an event about WebAI and High-perf Web APP
… the high-perf baseline CG was set up later after this event
… and in 2025 Sep, the WebEvolve annual conference was helded in Hangzhou
… it was a 2-day event
… Day 1 was about high-perf App
… Day 2 was about AI
… 26 talks, 100 attendees on site and 15000 obsevers online
… in Sep 2025, we attended the INclusion conference 2025 as well
… talked about security and identity by Simone
… right before TPAC, Seth attended the World Internet Conference 2025 online and gave a talk about next generation of identity
… so far there are 62 participants
… 3 co-chairs from Alibaba, Huawei and Interl
… there are comminication channels for the group
… welcome to join us!
… this group speaks both Chinese and English
AnQing: thank you Xueyuan for the intro!
Mokoto san: a question to ask
… soem specs are very language and cultuaal depending
… what did the JP do, we translate and articipate in international standards work
… what happends if internaitonal standards are biased
… in the past I created XML Japanese Application Profile!
… there were some issues related to JP specific
… I was asked to generate some issues
… it was a member submittion to W3C, W3C members paid attention to it
… in the case of JLreq
… it addresses the JP layout requirements
… the W3C groups follow that doc
… it is a very good example
… it is a requirement doc only, not a W3C spec
… in the case of CSS, ppl are aware of the cultural and language specific issues
… I assume it's the case for other languages in CSS WG
… but not for other groups maybe
… so JP is in a difficult position, if we just translate W3C spec
… if we create a diff one such as WCAG, ppl say we fork another standards
… would like to know what you guys are thinking about it
Martin: I think it's important what you mentioned
… the WCAG WG would appriciate feedback like this
… maybe the HR will address it
… I think W3C community will pay attention to is
Mokota san: not sure if it would work this way
… e.g. I filed an issue in I18N
Fuqiao: as someone from the I18N group, I would say we do pay attention to the I18N issue
… but not sure what happended for WCAG2
Fuqiao: in China, we are lack of research in some layout topics
… not so much papers on this topic
… if we need to meet the guidelines, it sohld be led by reseach
Mokota san: I read some books and papers on a11y, diff ppl say diff things
… hard to reach consensus
Mokota san: what do you CHinese ppl do on W3C reoomendations?
Lola: what does JP do now?
MOkota: we do not create modifications
<Zakim> lola, you wanted to react to martinA
MOkota: now I am thinking about modifications
angel: I want to respond to Mokoto san about what we do in China
… the Chinese community launched this accessibility national standard working group, and invited W3C people to join the group
… lookin into WCAG 2.2
… with the goal to harmonize the standard
… We adopted about 90% of the WCAG literal translation as-is. The remaining 10% involved modifications, primarily by lowering the threshold for some success criteria.
<Zakim> Dingwei, you wanted to ask any specific cultural issues
DingWei: I think we have comprehensive ways to adopt internaitonals standards
… in China, ISO and IEC have they way into local standands
… is it possible if we could start a repo to discuss it
… we have good methodology to that
<Zakim> martinA, you wanted to comment on another different topic (about ecosystem in China)
Martin: we had an interesting project on this
… A project on credentials in sports, and standardizing the documents (verifiable credentials). Started in Europe, but global outreach. If you have any contact in China, please let me know.
Martin: we have started a CG in W3C
… hopefully you can onboard some W3C members on this
Chunming: the IG do not publish W3C Recs
… it only provides a forum to discuss ideas
… hopefully we could incubute new groups for further standards work
… this might be the way to influence the standards work
… currently we have very good local events to incuate
… there are also good W3C mechamisms to incubate ideas as well
… we could try to link to them and work togehter with them
Anqing: very good discussion
… we can keep this in mind
… the scope of this group is a bigger topic
… we can talk about it later
… will hand the meeting to my co-chair WanLin
<PaolaDM> which are the authorities in handling digital identities in eu and china?
Miniapps standards update
Anqing: let's hear it from Martin the Co-chair of Miniapps WG
Slideset: https://
Anqing: will intro the the Mniapps standards activities
… there is an ecosystem of Miniapps in China, as well as in the world
… if you search the Miniapps outside of China in 2016
… there was limited content you could find
… but now the Miniapps are growing outside of China as well
… we can find various Miniapps used in many parts of the world
… most of the activities are in Asia
… some are in EU
… there are something similar like Light apps
… some are running on the OS
… some are like widget
… these miniapps can run on various platforms, such as native apps, OSs, or other forms
… some can run within a broswers as well
… there are similarities between the super apps and miniapps
… the main challlenges we have in the WG is how to enhance the UX
… wrt Development: how to code my MiniApp for n+1 SuperApps
… we have specific elements, APIs, ..
… wrt Publication: how to deliver it on different platforms (i.e., registration, validation, maintenance)
… a big pain point for developers
… and also not easy wrt versioning and maintenance...
… we began the work in the Chinese IG with a white paper on Miniapps in 2020
… in 2021, we launched the Miniapps WG
… we started to gain attention from the community
… the deliverables of this group got mentioned in @@3 in EU
… we involved the whitepaper bu adding more content into it
… e.g. practices in China, tech structure
… in TPAC2023, we started the review and found some security issues
… we did some fix work and continued the conversatoin
… and then we slowed down a bit
… now we have 6 docs in the WG
… about Miniapps lifecycle, addressing, widget req, packaging, and components
… and also working on address some issues, e.g. origin related security isseus
… we did an anlysis of the Miinapps API gap
… here is a summary
… some main stream super apps are analyized here
… we have a long list of APIs used by the super apps
… trying to find tou if there is a standards in W3C to support the requirements MiniAPps has
… about the market elements, it is quite similar, since based on HTML
… there are some specific element depending on platfrom which can be replaced by web componets or standard HTML element
… some current practices can be replaced by W3C standards
… we had the discussionabout mobile web engines
… most of the dicussions was about Miniapps
… e.g. how to build engine for Miniapps
… how to represent the Miniapps with webview
… it is possible to create lighter engines for Miniapps
… these engines can support other web-based apps
… thah is also relatd to the high-perf baseline CG work
… that group wants to create more efficient approach for cross-platform mobile app
[Martin demos a Miniapp-like apps on mobile]
… the Miniapps are light, can leverage the Web
… they are really cross platforms
… there are many opportunities
… MiniAPps are promising in Asia, but still unknown in other parts of the world
… but the interests is growing out side of Asia
… the merits of Web standards is that Web standards bring security, internationalization, accessibility by default
… some platforms might have these considerations, but not all
… we will continue the Miniapps work
… will do the recharting
… and add new work
… are ready to discuss the new requirements
… have new tech to consider
… this new isolated web app is something to consider
… e.g. packaging, distribution through HTTP
… we have developers interested in this
… we are also considering to mrge the work with High-Performance Engines work, maybe we could work together on dedicate and performant web engines
… welcome to join our meeting tomorrow
… happy to have more feedback
Lola: thank you for the informative talk
Slideset: http://
Lola: 2 questions
… what's the time for the meeting you mentioned tomorrow?
Martin: 9 am
<xfq> tomorrow's meeting: https://
Lola: you mentioned aobut the struggle for develpers to build for multi platforms
… do you foresee leveraging web stnadards can help with the issue?
Martin: we are still analizing it
… there are some distribution ways, one of them is HTML5
… some challgens with it
… it can be solve by 3rd party framework
… whcih can bring cross-platfrom capability
… it should be backed by the real implementers
… need the suport from the commuity in the end
… it's feasible to do it with web tech
DingWei: thank you martin for the talk
… this WG has been doing great job
… and also we identited some issues
… seems we should rethink about the recharter about the adding new tech
… this Miniapps idea is from China, and not gets more interests from the globe
… there are some similar solutions such as PWA
<chunming> IWA: https://
DingWei: we want to talk more about rethink about global adoption
… to my understanding, from architecture level, PWA is like MIniapps
… from implementation level, there are engines to support
… maybe time to think about the how structure
… that is why I think we need to have conversations with more ppl from other groups
Martin: I agree
… Miniapps is a new topic
… for developers outside of China
… they see the possibilites and have interests
… I heard it from a group of developers who are interested to try it
<Dingwei> https://
Martin: light engines for delivering something light to the users
Wanlin: thank you Martin!
<Dingwei> welcome to join in this Friday talk, engaging with IWA people who will join too...
Wanlin: happy to hear the progress
Web AI Agent Rendering Containers - Dynamic
[This talk is in Chinese]
Hai Rao: my talk today consists of four parts
… 1st, the Core Trend: AI is shifting from generating content to generating UIs… 2nd, the Generative UI Challenges: evaluating the three routes "Markdown, DSL, and Raw Web"
… 3rd, the Proposal: a Web AI Agent subset standard
… 4th, the Conclusions
… In 2025, "Agent" is the key concept for new AI applications. dimensions.
… Large language model applications are shifting across three key dimensions.
… This is more than a tech upgrade; it's redefining how humans and AI will collaborate in the future.
… AI Agents are advancing from simple dialogue to generating reports
… and now to interacting with the physical world and performing complex tasks.
… AI Agents are evolving from following preset rules to autonomously decomposing complex goals, planning, and using tools to accomplish them.
… The AI interaction is evolving beyond text boxes to provide a native experience, even generating dynamic UIs on demand.
… As the interaction paradigm of AI Agents evolves, their rendering layer has also progressed through three distinct stages.
… These advancements are driven by the growing capabilities of LLMs and Agents, alongside the increasing complexity of application scenarios.
… On one hand, users are no longer satisfied with pure text-based communication with LLMs.
… On the other hand, products aim to expand the generative paradigm from "producing text and content" to "generating UIs and applications."
… We have observed that the rendering needs of AI applications have transitioned through three stages:
… 1) streaming text and Markdown, 2) card-based layouts within workflow paradigms, and 3) generative UI.
… Stage 1: Streaming Text & Markdown
… Most LLMs adopt Markdown for rich text output due to its clear syntax and lower ambiguity.
… Early on, when LLM capabilities were limited, they could reliably generate Markdown but not necessarily HTML.
… Economically, Markdown consumes fewer tokens—HTML requires 5-6 times more for similar rendering.
… This approach separates concerns, allowing LLMs to focus on logical markup like bolding or quoting without styling details.
… Stage 2: Costom Cards under the Workflow Paradigm
… Early AI apps advanced to the "Workflow" paradigm, featuring modular splitting with pre-built card templates and standard interfaces.
… This required dynamic loading beyond code splitting, plus a communication mechanism for state synchronization between interactive cards and the main conversation.
… Stage 3: Agentic AI and the "Generative UI" Paradigm
… Agentic AI completes tasks autonomously, but this breaks the fixed 'intent-to-card' binding, forcing interfaces back to basic Markdown.
… The solution is Generative UI (GenUI), which uses standardized components to let the Agent dynamically assemble the right interface when needed, rather than selecting pre-set layouts.
… "Agent" is now the core of AI applications, with LLMs transitioning across three key dimensions, redefining human-computer collaboration.
… the three Routes of Generative UI: Enhanced Markdown, Proprietary DSLs, Raw Web (HTML/JS/CSS)
… Agents evolve from dialogue to physical interaction via tools and context protocols.
… Agents advance from preset paths to autonomous goal decomposition and execution.
… AI interaction progresses beyond text to dynamic, on-demand UI generation.
… The "Raw Web" Fallacy: Compounded Challenges
… Task evolution: from dialogue to reports, now interacting with the physical world via tools.
… Autonomy growth: from preset paths to self-directed goal decomposition and tool usage.
… Interface shift: beyond text boxes to dynamic, on-demand UI generation.
… The “Raw Web” Fallacy: Infinite Canvas vs. Limited Frame
… The web is an "infinite canvas" of evolving standards and APIs
… User devices are "limited frames" with specific technical constraints
… Asking AI to generate Raw Web code compounds technical challenges
… A "Trilemma": We Need a Fourth Way
… AI Agents evolve from dialogue to physical interaction via context protocols and tools.
… Their autonomy grows from preset paths to self-directed planning and tool usage
… Interaction advances beyond text to dynamic UI generation for native experiences.
… The Proposal
… RFC: Should a subset standard for Web AI Agent be established:
… DSL is a temporary fix that causes fragmentation.
… We need a lean, cross-platform web AI standard where LLMs generate standardized subsets instead of raw web or custom DSLs.
… The Solution: A Standard Subset for Web AI Agents
… Conclusions: build Web AI rendering containers
… Conversation Container (Native, MiniApp, HTML, PC), Container Rendering Layer, Runtime, Communication Layer
… Generative AI now creates interfaces and apps, dynamically built by AI Agents at runtime.
… This shift creates web friction in performance and security, demanding new standards.
… Developers should adopt AI Web Containers to deliver smarter user experiences.
… We need open, secure AI rendering standards for cross-platform intelligent applications.
Wangming: Thank you for this presentation. my question is about which groups that we could bring this topic to?
Xueyuan: we have an AI Agent Protocal CG and an upcoming Web & AI IG
… this could be a good topic for them
SM2 & SM4 vs WebCrypto
[This talk is in Chinese]
Slideset: https://
Zhilei Zhang: I'm from Qihoo 360.
… Today I'll share insights about China's commercial cryptographic algorithms SM3 and SM4… along with practical implementations for supporting these algorithms in WebCrypto interfaces.
… My presentation consists of three main parts: 1st, a brief introduction to SM3 and SM4 algorithms;
… second, their implementation in WebCrypto interfaces
… and finally, our future expectations.
… SM3 Hash Algorithm … The SM3 is a hash algorithm with strength comparable to SHA-256
… designed by Academician Wang Xiaoyun and other Chinese experts.
… It was released as an industry standard in 2012
… elevated to national standard in 2016, and adopted by ISO as an international standard in 2018.
… Currently, it's primarily used in TLCP protocols, digital signature hash operations, and various data integrity verification scenarios.
… SM4 Symmetric Encryption Algorithm
… SM4 is a symmetric encryption algorithm with strength equivalent to AES.
… It became an industry standard in 2012, a national standard in 2016, and was adopted by ISO as an international standard in 2021.
… It's mainly used in TLCP protocols and data encryption scenarios.
… Regulatory Requirements in China
… In China, specific cryptographic requirements for information systems are established through multiple frameworks:
… Cryptographic evaluation requirements in GB/T 39786 specify that cryptographic technologies must comply with national and industry standards
… Critical infrastructure protection mandates commercial cryptography usage through dedicated regulations
… The classified protection scheme requires network operators to use commercial cryptography for cybersecurity protection
… For browsers specifically, the 2023 government procurement standards explicitly require support for commercial cryptographic algorithms
… that meet national cryptographic management requirements.
… WebCrypto Implementation Status
… Currently, WebCrypto's digest method supports four hash algorithms (SHA-1, SHA-256, SHA-384, SHA-512) but lacks SM3 support
… despite SM3 having comparable strength to SHA-256.
… Similarly, while WebCrypto's encrypt method supports three AES modes (CBC, GCM, CTR)
… all compatible with SM4, the SM4 algorithm identifiers remain undefined in the standard.
… However, China's State Cryptography Administration has published GM/T 0087-2020
… standardizing SM3 and SM4 usage in WebCrypto interfaces.
… The implementation methods align with existing WebCrypto approaches
… simply replacing identifiers with "SM3" and "SM4-CBC".
… Implementation Progress
… Open-source communities have gradually adopted these algorithms:
… Linux kernel introduced SM3 in 2017 and SM4 in 2018
… OpenSSL has supported both since version 1.1.1, with latest versions featuring Intel and ARM instruction set acceleration
… Other cryptographic libraries have also incorporated support
… 360 has implemented SM3 and SM4 in WebCrypto based on OpenSSL's implementation.
… Performance tests show significant improvements:
… SM3 via WebCrypto is 100-300x faster than JavaScript implementations, while SM4 shows 30-40x performance gains.
… Certification and Adoption
… In China, products including browsers require commercial cryptography product certification for use in critical infrastructure and classified protection scenarios.
… Browser certification mandates support for GM/T 0087 and related standards.
… Currently, approximately a dozen browsers have obtained this certification.
… With multiple browsers already implementing SM3 and SM4 in WebCrypto interfaces
… covering hundreds of millions of users, we hope to formally add SM3 and SM4 algorithms to the WebCrypto standard
… providing clear implementation guidance for browser vendors.
Simone: we could discuss this with the WebAppSec WG and bring the attention to the group
… it is good to put in the agenda item for the next one (calendar here)… Security IG is preparing a guideline document about crypto in standards
… and as we wrote about national standardsbodies, some cryptographers complained (and this can be an issue if SM family is from a national body)
Simone: to mitigate this, it can be good to raise the issue also in IETF CFRG
… Next joint meeting with IETF is in Shenzhen in March (IETF 125)… as a liaison between W3C and IETF we can help
[No more questions on Web Crypto]
Wanming: We have covered all the topics for today
… open discussion now
… Please raise your comments to the group
Phil: I work for GS1,
… Chair of Verifiable Credentials
… you asked for a method for verifiable credentials
… VC is composed of data models, security mechanisms (emmbeded and eveloping)
… If there is a interest in this group for creating a new security mechanism based on SM3 or SM4 algorithms
… I want to understand your proposal
xfq: [translating in Chinese]
wanming: I'll come back to you offline
Zhilei: see the SM3 issue w3c/
PhilA: I want to know about the standard, if there is a Chinese standard...
… Is this Chinese Web group a CG?
xueyuan: it's an IG
PhilA: So cannot produce specs
<xfq> SM3 is also in ISO/IEC 10118-3:2018
<xfq> SM4 in ISO/IEC 18033-3:2010/Amd 1:2021
PhilA: If you want to put this requirement into the VC WG re-chartering
… This is the right moment to do it
… This could be created within the VC WG, specs, test suites, editing.
… Tomorrow is the meeting
<phila> phila: Details of the VCWG meeting tomorrow are at https://
wanming: any comments?
xueyuan: We want to make the Chinese IG better support members and the community need
… and we want to know how we can improve the communication with other groups, and bring new topics into our IG
PhilA: Speaking as my role in GS1 with barcodes, VCs...
… GS1 is multiple orgs, including several in China (in various Chinese provinces)
… Thanks for the Chinese translation of VCs.
… We talk about cross-country trade documentation, it's about exchange of technology
… Today there is a breakout session. It's about global exchange of credentials
… I would be delighted to have more communication to achieve this cross-country data exchange
xueyuan: we are happy to support by organizing events on relevant topics
PhilA: GS1 does QR codes, but many other things.
xueyuan: Any other comment how we do the communication with the Chinese ecosystem
Bobby: no comments
Wanming: Thanks for your time today.
… Lunch time
[meeting adjourned]