14:55:05 <RRSAgent> RRSAgent has joined #webmachinelearning
14:55:09 <RRSAgent> logging to https://www.w3.org/2025/09/18-webmachinelearning-irc
14:55:09 <Zakim> RRSAgent, make logs Public
14:55:10 <anssik> Meeting: WebML CG Teleconference – 18 September 2025
14:55:10 <Zakim> please title this meeting ("meeting: ..."), anssik
14:55:17 <anssik> Chair: Anssi
14:55:31 <anssik> Agenda: https://github.com/webmachinelearning/meetings/blob/main/telcons/2025-09-18-cg-agenda.md
14:55:36 <anssik> Scribe: Anssi
14:55:39 <anssik> scribeNick: anssik
14:55:48 <anssik> Present+ Anssi_Kostiainen
14:55:59 <anssik> RRSAgent, draft minutes
14:56:00 <RRSAgent> I have made the request to generate https://www.w3.org/2025/09/18-webmachinelearning-minutes.html anssik
14:59:34 <anssik> Present+ Joon_Park
15:00:01 <anssik> Present+ Thomas_Steiner
15:00:18 <anssik> Present+ Tarek_Ziade
15:00:48 <anssik> Present+ Khushal_Sagar
15:00:58 <anssik> Present+ Alex_Nahas
15:01:24 <Leo> Leo has joined #webmachinelearning
15:01:32 <kush> kush has joined #webmachinelearning
15:01:35 <tarek> tarek has joined #webmachinelearning
15:01:36 <kush> present+
15:01:37 <brwalder> brwalder has joined #webmachinelearning
15:01:54 <brwalder> present+
15:01:56 <Leo> present+
15:01:59 <anssik> Present+ Brandon_Walderman
15:02:20 <anssik> Present+ JD
15:02:25 <Ehsan> Ehsan has joined #webmachinelearning
15:02:34 <anssik> Present+ Andrew_Nolan
15:02:44 <anssik> Present+ Leo_Lee
15:03:39 <anssik> RRSAgent, draft minutes
15:03:40 <RRSAgent> I have made the request to generate https://www.w3.org/2025/09/18-webmachinelearning-minutes.html anssik
15:03:46 <nerding-io> nerding-io has joined #webmachinelearning
15:03:58 <anssik> Anssi: first, please welcome Markus Tavenrath from NVIDIA to the WebML Community Group!
15:04:42 <anssik> Markus: I'm curious how Web and ML is evolving, NVIDIA is interested in all things going on in this space, decided to join both the groups WG and CG
15:05:31 <anssik> ... the web is a great place to deploy ML, somewhat limited to fully exploit modern hardware, but want to help unlock the full potential
15:07:07 <anssik> ... Joon from Target Corp, engineering director, we want to provide input how consumers would be using technology developed in this group
15:07:33 <anssik> ... worked with the Chrome team, testing new APIs e.g. Built-in AI APIs
15:08:18 <anssik> Joon: WebMCP API is of interest to our senior leaders
15:09:02 <anssik> Anssi: this is the new meeting time for the WebML Community Group to better support the AMER geo during the WebMCP ramp-up phase
15:09:16 <anssik> ... this meeting is interleaved with the Working Group meeting at this familiar time slot.
15:09:25 <anssik> ... we're using IRC for this CG meeting to make use of custom IRC bots, recording resolutions etc.
15:09:39 <zkis> zkis has joined #webmachinelearning
15:09:42 <anssik> ... I consulted W3C staff on the use of AI-powered tools to transcribe, summarize, and record meetings
15:09:42 <stalgiag> stalgiag has joined #webmachinelearning
15:09:47 <anssik> ... it is allowed if permissions from all participants is granted
15:10:09 <anssik> ... in our earlier experimentation, the transcript had errors due to domain specificity, while the summary was more useful
15:10:12 <anssik> ... we can continue test AI-powered meeting tools but won't share output publicly to avoid misinterpretation
15:10:27 <zkis> present+ Zoltan_Kis
15:10:33 <anssik> Topic: WebML CG Charter update
15:10:38 <anssik> gb, this is webmachinelearning/charter
15:10:39 <gb> anssik, OK.
15:10:44 <anssik> -> Call for review https://lists.w3.org/Archives/Public/public-webmachinelearning/2025Aug/0005.html
15:11:01 <anssik> Anssi: we initiated the review of the WebML CG Charter update a month ago to make WebMCP API a new WebML CG deliverable
15:11:07 <anssik> ... thank you for your review and support
15:11:17 <anssik> ... for the record, we received 10 support votes on the PR by 2025-09-18 via +1s and no objections
15:11:27 <anssik> ... this charter update passes with unanimous support, the updated charter is operational on on 25 September 2025
15:11:53 <anssik> ... during the review period we received two new explainers that supplement the main explainer:
15:12:00 <anssik> -> WebMCP for Service Workers (merged) https://github.com/webmachinelearning/webmcp/blob/main/docs/service-workers.md
15:12:11 <anssik> -> Declarative API (in review) https://github.com/webmachinelearning/webmcp/pull/26
15:12:12 <gb> https://github.com/webmachinelearning/webmcp/pull/26 -> Pull Request 26 add explainer for the declarative api (by MiguelsPizza)
15:13:17 <anssik> Anssi: I consider both the explainers to fall within the scope of the WebMCP API incorporated in the revised charter with a provision these extensions comply with the following statements in the Scope of Work:
15:13:21 <anssik> ... - "designed for human-in-the-loop workflows where the user maintains visibility and remains in control"
15:13:25 <anssik> ... - "Each specification should detail all known security and privacy implications for implementers, Web authors, and end users"
15:13:35 <anssik> ... questions or comments?
15:13:43 <anssik> q?
15:13:48 <anssik> q+ to ask a question
15:13:52 <anssik> ack anssik
15:13:52 <Zakim> anssik, you wanted to ask a question
15:14:11 <AlexN> AlexN has joined #webmachinelearning
15:14:14 <anssik> Topic: WebMCP API
15:14:19 <anssik> gb, this is webmachinelearning/webmcp
15:14:19 <gb> anssik, OK.
15:14:55 <anssik> Subtopic: Use cases
15:15:00 <anssik> Anssi: issue #27
15:15:01 <gb> https://github.com/webmachinelearning/webmcp/issues/27 -> Issue 27 Use-cases targeted by WebMCP (by khushalsagar)
15:15:37 <anssik> ... broadly speaking, for WebMCP there could be two broad use case categories:
15:15:45 <anssik> ... 1) Automation, non-human in the loop
15:15:50 <anssik> ... 2) User and Agent sharing the UI, aka the human in the loop
15:15:56 <anssik> ... it seems we have an agreement to focus on the latter use cases where human is in the loop
15:16:00 <anssik> ... this is also aligned with our charter scope of work
15:16:04 <anolan> anolan has joined #webmachinelearning
15:16:11 <anssik> ... any other comments?
15:16:27 <anssik> +1 from Jason
15:16:29 <thelounge> thelounge has joined #webmachinelearning
15:16:51 <scheib> scheib has joined #webmachinelearning
15:17:50 <anssik> Khushal: I assume there was general consensus we focus on human in the loop use cases
15:18:33 <anssik> Alex: I agree, MCP-B is human in the loop app first, there's a lot of products that are doing browser automation via OCR-style approach and clicking
15:18:44 <anssik> ... when we figure out security story we can revisit this later
15:18:45 <anssik> q?
15:19:56 <anssik> RESOLUTION: WebMCP focuses on human in the loop use cases initially.
15:20:01 <Joon> Joon has joined #webmachinelearning
15:20:04 <anssik> RRSAgent, draft minutes
15:20:05 <RRSAgent> I have made the request to generate https://www.w3.org/2025/09/18-webmachinelearning-minutes.html anssik
15:20:11 <Joon> Joon has joined #webmachinelearning
15:20:14 <anssik> Subtopic: Core design principles
15:20:20 <anssik> Anssi: issue #25
15:20:21 <gb> https://github.com/webmachinelearning/webmcp/issues/25 -> Issue 25 Core design principles for WebMCP (by khushalsagar)
15:20:46 <anssik> ... MCP defines 2 layers for how server-client communication is setup
15:21:03 <anssik> ... data layer - request/response definition for initialization
15:21:05 <anssik> ... transport layer - how messages are exchanged
15:21:25 <reillyg> present+ Reilly_Grant
15:21:33 <anssik> ...  proposal for WebMCP to use the "SDK option":
15:21:33 <anssik> ... data layer - browser translate MCP client <-> WebMCP server
15:22:11 <anssik> Khushal: wanted to ensure Msft's and Google's initial proposals align on this topic
15:22:53 <anssik> ... primitives are the basic data objects e.g. tools, WebMCP reuses primitives, not data layer
15:23:31 <anssik> ... I propose we should go with the "SDK option"
15:24:31 <anssik> Brandon: if I got the proposal correct, we're supportive of "SDK option"
15:24:43 <stalgiag> present+ stalgiag
15:26:01 <anolan> q+
15:26:32 <anssik> Andrew: it seems important for WebMCP to be able to work without an official MCP server
15:26:48 <anssik> ... these could be generic actions that open new possibilities for the API
15:27:00 <anssik> ack anolan
15:28:03 <anssik> Anssi: "SDK option" pros:
15:28:07 <anssik> ... - decoupling WebMCP from MCP version
15:28:12 <anssik> ... - allows for different interaction patterns between WebMCP and MCP (see #21)
15:28:13 <gb> https://github.com/webmachinelearning/webmcp/issues/21 -> Issue 21 Elicitation (by bwalderman)
15:28:16 <anssik> ... - API ergonomics improvement, allow img and video element output
15:28:21 <anssik> ... - declarative API easier
15:28:58 <anssik> Jason +1
15:29:03 <anssik> Brandon +1
15:29:07 <anssik> Alex +1
15:29:12 <anssik> RESOLUTION: WebMCP to provide a layer of abstraction between the browser and MCP client aka "SDK option".
15:29:26 <AlexN> AlexN has joined #webmachinelearning
15:30:45 <anssik> Subtopic: Bikeshedding the global name
15:30:53 <anssik> Anssi: issue #24
15:30:53 <gb> https://github.com/webmachinelearning/webmcp/issues/24 -> Issue 24 Bikeshedding the global name (by bwalderman)
15:31:14 <anssik> Anssi: window.mcp is gaining support
15:32:10 <anssik> Thomas: wanted to bring a point that Apple brought up feedback that web is something that will be here for tens of years to come and we should strive for neutral names for globals that stand the test of time
15:32:13 <kush> q+
15:32:22 <anssik> ack kush
15:32:53 <reillyg> q+
15:32:58 <anssik> Khushal: I hear you, MCP is parallel to GL in terms of naming, we don't know how long this primitive lives
15:33:09 <anssik> ack reillyg
15:33:40 <anssik> Reilly: adding to window object makes it part of global
15:33:57 <anssik> ... has a decent likelyhood to collide with JS obfuscators
15:34:28 <anssik> ... mediating between the agent and user, putting this on navigator would be preferred
15:35:28 <anssik> Jason: wanted to mention that as Reilly mentioned, agent registry or a collective object skirts the issue for window.agent where you may have multiple agents down the road
15:35:30 <anssik> q?
15:36:09 <anssik> Anssi: does this issue block some experimental implementation?
15:36:30 <anssik> Khushal: there's an experimental implementation currently that depends on our naming discussion
15:37:13 <brwalder> q+
15:37:21 <anssik> Khushal: is there consensus to put this on navigator object?
15:37:30 <anssik> ack brwalder
15:38:05 <anssik> Brandon: we got internal feedback that navigator is associated with the entire browser, not specific window or page, thus going with window initially
15:39:00 <anssik> Reilly: document object would make more sense than window
15:40:32 <anssik> Jason: parallels to permissions interface?
15:40:35 <anssik> Reilly: exactly
15:40:56 <anssik> Anssi: proposal to drive consensus via GH issue discussion in the issue #24 to converge on naming and where to hang it off of: window, document, navigator
15:40:56 <gb> https://github.com/webmachinelearning/webmcp/issues/24 -> Issue 24 Bikeshedding the global name (by bwalderman)
15:41:14 <anssik> Subtopic: API design
15:41:18 <anssik> Anssi: issue #15
15:41:18 <gb> https://github.com/webmachinelearning/webmcp/issues/15 -> Issue 15 API design (by bwalderman)
15:41:25 <anssik> ... Brandon outlines two API design approaches:
15:41:29 <anssik> ... 1) An object with registerTool(options) and unregisterTool(name) methods
15:41:37 <anssik> ... 2) A single provideContext method which takes a collection of tools in a single call.
15:41:56 <anssik> ... Alex' implementation and Google's Script Tools API proposal used (1),  while the current WebMCP proposal uses (2)
15:42:17 <anssik> ... Jason asked about multi page web apps and persistence, his implementation persists tools to local storage, there seems to be agreement Service Workers could solve this
15:42:41 <anssik> Jason: I'm not advocating for local storage for the spec, it is an implementation detail
15:42:53 <kush> q+
15:42:54 <brwalder> q+
15:42:58 <anssik> ack kush
15:43:23 <anssik> Khushal: we have a set of tools, the question is can I addFoo and removeFoo or do it via an entire list
15:43:50 <anssik> Anssi: do we have consensus on this issue to stay with the current proposal (2)?
15:44:12 <brwalder> q+
15:44:17 <anssik> Anssi: what are the obstacles in supporting both (1) and (2) and do we need more experience to understand the developer ergonomics improvement better?
15:44:54 <anssik> Alex: for dynamic web sites, React-based (1) is beneficial, I will try to polyfill to provide implementation experience
15:45:05 <reillyg> q+
15:45:08 <anssik> Khushal: we could have both (1) and (2)
15:45:32 <anssik> ack brwalder
15:45:55 <anssik> Brandon: in Chromium our plan is to implement both (1) and (2) and allow developers to experiment and gather feedback
15:46:39 <anssik> ... I actually pitched (2) with React in mind, with some state you can compute what is available and batch send them
15:47:51 <anssik> Khushal +1
15:47:56 <anssik> Brandon +1
15:47:59 <anssik> Alex +1
15:48:03 <anssik> RESOLUTION: Continue use with the provideContext design and explore adding support for registerTool(options) and unregisterTool(name) to complement the API informed by implementation experience.
15:48:14 <anssik> Subtopic: Declarative API
15:48:20 <anssik> Anssi: issue #22 and PR #26 for the explainer
15:48:20 <gb> https://github.com/webmachinelearning/webmcp/issues/22 -> Issue 22 Declarative API Equivalent (by EisenbergEffect)
15:48:20 <gb> https://github.com/webmachinelearning/webmcp/pull/26 -> Pull Request 26 add explainer for the declarative api (by MiguelsPizza)
15:48:40 <anssik> ... a proposal to enable web pages to expose tools via HTML custom tool-* attributes
15:48:49 <anssik> ```
15:48:49 <anssik> <form action="/todos" method="post" tool-name="add-todo" tool-description="Add a new todo item to the list">
15:48:49 <anssik>   <input type="text" name="description" required tool-prop-description="The text of the todo item">
15:48:50 <anssik>   <button type="submit">Add Todo</button>
15:48:50 <anssik> </form>
15:48:50 <anssik> ```
15:49:10 <anssik> Anssi: this proposal broadens the scope of the API to content creators
15:49:26 <anssik> ... declarative has a significantly broader reach than imperative API among developers, expands to content authors
15:49:43 <tomayac> q+
15:49:49 <anssik> ... learnings from Web Components suggest an imperative-only API without a declarative is problematic
15:50:09 <anssik> ack tomayac
15:50:16 <anssik> ack reillyg
15:51:40 <anssik> Thomas: I'm wondering what's implementers expectation for JS-rendered apps?
15:52:08 <anssik> ... do you expect people to harvest declarative instructions to execute as JS, or should there be a separate way, e.g. <noscript> style?
15:52:09 <anssik> q?
15:52:40 <anssik> Alex: what do you mean by expectations, Thomas?
15:53:14 <anssik> Thomas: what will happen in the ecosystem, e.g. open graph instruction, people can deliver them via JS, you have to have them in HTML delivered from the server
15:53:29 <anssik> ... would the same be true for declarative instructions for this API
15:53:51 <anssik> Alex: declarative is complementary to <form>s
15:54:06 <kush> q+
15:54:21 <anssik> ... for React usage, JS API provides the needed flexibility
15:54:43 <anssik> ... my demo was not expensive to implement
15:55:10 <anssik> ... couple of issues noted in the issue, how to handle elicitation
15:55:20 <anssik> ... not sure how people want to handle it
15:55:55 <anssik> ... another is expectation for the web server, WebMCP currently is JSON-RPC protocol and web apps that only serve HTML, do we have them add a new route for JSON?
15:55:56 <anssik> q?
15:56:18 <anssik> Thomas: one alternative way to deliver would be via .well-known URL with declarative tools the page support?
15:56:45 <anssik> q?
15:57:57 <anssik> ack kush
15:58:33 <anssik> Khushal: we don't know all the use cases for tools, some might be tied to HTML elements, I wonder if we see agent specific tools that make no sense to have UI associated
15:58:54 <anssik> ... I like the direction of this issue, wanted to identify edge cases
15:59:17 <anssik> ... good to drill down into details of declarative to be able to ship it
15:59:32 <anssik> RRSAgent, draft minutes
15:59:34 <RRSAgent> I have made the request to generate https://www.w3.org/2025/09/18-webmachinelearning-minutes.html anssik
16:00:14 <anssik> Alex: PR welcomes review and comments
16:01:05 <kush> q+
16:01:24 <anssik> Subtopic: To triage or not to triage
16:01:28 <anssik> Anssi: issue #18
16:01:29 <gb> https://github.com/webmachinelearning/webmcp/pull/18 -> MERGED Pull Request 18 Convert SVG to PNG for easier reading in light/dark modes (by bwalderman)
16:01:33 <anssik> ... I wanted to solicit feedback on the group's preference for WebMCP API spec issue triage practices
16:01:38 <anssik> ... thanks Khushal for sharing your CSSWG experience
16:01:46 <anssik> ... I'm supportive of documenting resolutions in the minutes now that the work has ramped up
16:01:58 <anssik> ... for labels, please provide your suggestions in the issue
16:02:04 <anssik> ... I adopted "Agenda+", how has CSSWG set up its triage GH permissions?
16:02:46 <anssik> Anssi: do all the CSSWG apply Agenda+ to issues?
16:02:58 <anssik> Khushal: I think everyone in the group
16:03:27 <anssik> Anssi: we can be liberal and allow all members to add labels
16:04:00 <anssik> Khushal: CSSWG has IRC-GH integration
16:04:48 <anssik> RRSAgent, draft minutes
16:04:50 <RRSAgent> I have made the request to generate https://www.w3.org/2025/09/18-webmachinelearning-minutes.html anssik
16:06:11 <anssik> Present+ Ehsan_Toreini
16:06:36 <anssik> Present+ Jason_McGhee
16:06:46 <anssik> Present+ JD_Fiscus
16:06:58 <anssik> Present+ Laura_Morinigo
16:07:25 <anssik> Present+ Markus_Tavenrath
16:07:39 <anssik> Present+ Stalgia_Grigg
16:07:49 <anssik> RRSAgent, draft minutes
16:07:50 <RRSAgent> I have made the request to generate https://www.w3.org/2025/09/18-webmachinelearning-minutes.html anssik
16:09:04 <anssik> s/Joon from/Joon: Joon from
16:15:18 <anssik> s/Alex'/Alex's
16:17:11 <anssik> s/implementers expectation/implementers' expectation
16:21:03 <anssik> RRSAgent, draft minutes
16:21:05 <RRSAgent> I have made the request to generate https://www.w3.org/2025/09/18-webmachinelearning-minutes.html anssik