14:02:28 <RRSAgent> RRSAgent has joined #epub-locators
14:02:28 <RRSAgent> logging to https://www.w3.org/2021/04/21-epub-locators-irc
14:02:39 <ivan> rrsagent, set log public
14:02:49 <ivan> Chair: wendyreid
14:02:50 <ivan> Date: 2021-04-21
14:02:50 <ivan> Meeting: EPUB 3 Working Group Virtual Locators Task Force Telco
14:23:07 <wendyreid> wendyreid has joined #epub-locators
14:35:42 <tzviya> regrets+
14:51:58 <wendyreid> zakim, start meeting
14:52:00 <Zakim> RRSAgent, make logs Public
14:52:00 <Zakim> please title this meeting ("meeting: ..."), wendyreid
14:52:17 <wendyreid> meeting: Virtual Locators TF Telco April 21, 2021
14:52:22 <wendyreid> date: 2021-04-21
14:52:26 <wendyreid> chair: wendyreid
14:55:01 <dauwhe> dauwhe has joined #epub-locators
14:57:40 <ronnie_> ronnie_ has joined #epub-locators
14:58:29 <BenSchroeter> BenSchroeter has joined #epub-locators
15:00:32 <dauwhe> present+
15:00:39 <dauwhe> Zakim, who is here?
15:00:41 <Zakim> Present: dauwhe
15:00:41 <Zakim> On IRC I see BenSchroeter, ronnie_, dauwhe, wendyreid, RRSAgent, Zakim, ivan, tzviya
15:00:59 <dauwhe> present+ wendyreid, ronnie_
15:01:39 <ivan> present+
15:01:47 <dlazin> dlazin has joined #epub-locators
15:02:00 <ivan> present+ brafy
15:02:05 <duga> duga has joined #epub-locators
15:02:05 <ivan> present+ brady
15:02:10 <ivan> present+ dlazin
15:02:10 <duga> present+
15:02:18 <BenSchroeter> present+
15:02:37 <dauwhe> present+ duga
15:03:52 <wendyreid> https://www.theverge.com/2021/4/17/22389519/google-feature-chrome-90-highlighted-links
15:03:54 <pilarw2000> pilarw2000 has joined #epub-locators
15:04:05 <pilarw2000> present+
15:04:06 <ivan> present+ pilarw2000
15:04:49 <ivan> present+ clapierre
15:04:54 <wendyreid> scribe+ ronnie_
15:05:10 <dauwhe> wendyreid: you just say who's speaking
15:05:28 <ronnie_> nickname: what they say
15:05:34 <dauwhe> ivan: you say the nickname, whatever the person said
15:05:43 <dauwhe> ... and if they keep talking
15:06:45 <dlazin> So for Wendy's link above, here is what a link-to-text anchor looks like: https://www.theverge.com/2021/4/17/22389519/google-feature-chrome-90-highlighted-links#:~:text=already-,available,-on
15:07:11 <BenSchroeter_> BenSchroeter_ has joined #epub-locators
15:08:13 <ronnie_> ben: direct-to-text anchor: Chrome 89 or later when you click it, and it appears in yellow. Only on Chrome.
15:08:35 <wendyreid> s/ben:/dlazin:/
15:08:42 <ronnie_> ben: Obvious problem of being non-stable, if the text changes, it doesn't work anymore.
15:08:42 <dlazin> https://www.macrumors.com/how-to/link-to-highlighted-text-webpage-chrome/#:~:text=Unfortunately%2C%20the%20highlight%20links%20that%20Chrome%20generates%20only%20work%20in%20Edge%20and%20Chrome%2C%20therefore%20users%20running%20other%20browsers%20won%27t%20see%20the%20highlighted%20text.%20However%2C%20they%27ll%20still%20be%20sent%20to%20the%20webpage%20in%20question%2C%20so%20the%20link%20isn%27t%20completely%20useless%20to%20Safari%20o[CUT]
15:09:00 <ronnie_> ben: This link is from a paragraph.
15:09:29 <ronnie_> ben: It's quite the URL but it works, but only in Chrome. It might become standardized but not yet
15:10:05 <ronnie_> Ivan: It gives the paragraph before in yellow.
15:10:27 <ronnie_> Ben: That might be because of 89; it was announced in 90.
15:11:04 <ronnie_> Ivan: Implementation details still needed. This reminds me of meta-structure to describe anchors in an old implementation WG spec.
15:11:14 <ivan> https://www.w3.org/TR/wpub-ann/
15:11:26 <duga> s/Ben/Dan/
15:11:33 <ivan> https://www.w3.org/TR/selectors-states/
15:11:48 <CharlesL> CharlesL has joined #epub-locators
15:13:39 <dauwhe> spec: https://wicg.github.io/scroll-to-text-fragment/
15:14:07 <ronnie_> Iva This is a metastructure that describes various ways you can select things in a resource ... text or something preceded or followed by something, or an SVG window, etc. Giving you a unified JSON structure to select it. And then in this node there's a nonstandard way to turn it into a URL. At least part has been implemented by Hypothesis, but not the URL part, which is just a thought exercise. In the web implementation WG we added annot[CUT]
15:14:24 <wendyreid> q?
15:14:36 <ronnie_> Ivan: you want to find a text preceded and followed by something.
15:14:48 <ronnie_> Wendy: we're not the only people working on this ...
15:15:01 <BenSchroeter_> q+
15:15:06 <wendyreid> ack BenSchroeter_
15:15:18 <dauwhe> Mozilla standards position on this: https://github.com/mozilla/standards-positions/issues/194
15:15:32 <wendyreid> https://docs.google.com/document/d/1y5JKcwlq1rvGYJ1HoLg0GF-NRiBsBNNZopGuQ0mgHUw/edit
15:15:53 <ronnie_> Ben: there's a page with examples of use cases. It's in the Google Doc above.
15:17:00 <ronnie_> Wendy: we looked at all the use cases and divided them into groupings; the bulk fit into one on specific locations - reference one or guide people to that, and offer affordances for a browser of OS to generate them efficiently.
15:17:13 <ronnie_> ... so we looked at CFI next
15:17:39 <ronnie_> ... couldn't get the list to stop numbering
15:18:06 <ronnie_> Ivan: Where do we go from here? We can't solve the general case for the whole web.
15:18:50 <ronnie_> Dan: It's problematic that so many others are working on this at the same time. Chrome is working on it and we have 3 W3 specs here, but none of them fully exists. A nice summary of where we are.
15:18:54 <duga> q+
15:19:07 <ronnie_> Ivan: We could say "use ePub CFI"
15:19:08 <wendyreid> ack duga
15:19:16 <dauwhe> https://github.com/w3ctag/design-reviews/issues/392
15:19:57 <ronnie_> Brady: 3 people are trying to solve a related but different problem. None of them are what we're trying to solve. Fix the fact that there are no anchors. For indexing or sharing info with people in a class - these aren't solved.
15:20:09 <BenSchroeter_> q+
15:20:15 <wendyreid> ack BenSchroeter_
15:20:18 <ronnie_> ... maybe we should wait, but that's worrisome. I do agree we don't need another spec that won't get adopted.
15:20:29 <ronnie_> Ben: Is a CFI reading system agnostic?
15:20:33 <ronnie_> Wendy: Yes
15:20:43 <duga> q+
15:21:09 <ronnie_> Ben: CFI is unreadable for humans and impossible for authors or indexers to use it because it's unreadable.
15:21:32 <ronnie_> ... It's perfectly possible for a reading system to use it internally. But who cares about having it as a standard then?
15:21:37 <wendyreid> ack duga
15:22:18 <ronnie_> Brady: CFI has some technical issues, some could be corrected by tooling. CFI's problem is there is no business case for it.
15:22:51 <ronnie_> ... Therefore, it hasn't been adopted. Product managers haven't said we really need elaborate indexes to work.
15:23:10 <ronnie_> ... Other things take precedence, not technical capacity.
15:23:31 <ronnie_> Pilar: That would be an argument for publishing the use cases
15:23:39 <ivan> q+
15:24:10 <ronnie_> Wendy: It depends on the framing. If you do this you can implement x-platform sharing, allowing book club members to reference the same quote
15:24:14 <duga> q+
15:24:17 <wendyreid> ack ivan
15:24:26 <ronnie_> Pilar: we need to angle it to how do they make money?
15:24:38 <ronnie_> Ivan: What is the implementation level of CFI?
15:24:47 <ronnie_> Brady: No one knows, not external.
15:25:25 <wendyreid> ack duga
15:25:43 <ronnie_> Ivan: If we have a use case document and we refresh the IDPF document as a note (not a recommendation). There's a hope that might push it to implementation. Or is it a fancy dream
15:26:12 <BenSchroeter_> q+
15:26:33 <ivan> q+
15:26:37 <ronnie_> Brady: No one knows how to write CFI, won't do it till the reading systems implement it, and the publishers won't yet ... It's a vicious cycle, hard to evolve till the pieces are implemented.
15:26:58 <wendyreid> ack BenSchroeter_
15:27:04 <ronnie_> ... we could do it, but there's no incentive to expose it in the UI and deal with the resulting hassles.
15:27:08 <CharlesL> q+
15:27:50 <ronnie_> Ben: Reading systems use deep linking currently to interface with management systems, from your syllabus you can deep-link to a specific paragraph.
15:28:16 <wendyreid> ack ivan
15:28:16 <ronnie_> ... Assumed they were using CFIs to do that, but maybe something else. There must be some implementation now that affords that use case.
15:28:36 <ronnie_> Ivan: I fully understand Brady's point, but I'm trying to be more optimistic.
15:28:46 <ronnie_> ... If the working grouDescription:https://mit.zoom.us/j/92468476393?pwd=Yk5ydXZqSTlla01PTkdlZVpuZ1lRdz09 Password: epub-wg Meeting ID: 924 6847 6393  https://irc.w3.org #epub-locators  present+ / q+
15:29:25 <ronnie_> If the working group refreshes the requirements and publishes it as a note, it draws attention to it and gives the message that it would be good to do it for these reasons.
15:29:26 <BenSchroeter_> https://www.imsglobal.org/spec/lti-dl/v2p0
15:29:47 <wendyreid> ack CharlesL
15:29:47 <ronnie_> Ivan: For the time being it's completely forgotten. It's the only thing we have.
15:29:51 <duga> q+
15:30:23 <ronnie_> Charles: Isn't ePub ACE already using CFIs for where the images are and the list of violations?
15:30:38 <ronnie_> ... Are we already doing it but don't have a way to access it?
15:30:43 <wendyreid> ack duga
15:31:25 <CharlesL> Here is an example from ACE -
15:31:25 <CharlesL> s9ml/chapter01/untitled_page_1.xhtml#epubcfi(/4/2[data-uuid-fb63c7f4d81a47f4a862351ff8e9d269]/8[data-uuid-70aed7f318c045ed8db490083a691b93]/8[data-uuid-e72dd28118af47dea3d815bcb0a85324]/2)
15:31:34 <ronnie_> Brady: One problem with CFI is you really want it when you don't control the content. If you control the content to generate an index, just insert spans with anchors used by the index entries.
15:31:51 <dlazin> q+
15:32:00 <ronnie_> ... mostly useful when I don't create the content. Most of uses don't matter if I don't have a universal identifier.
15:32:20 <wendyreid> ack dlazin
15:32:23 <ronnie_> ... How much use is there for CFI therefore?
15:32:52 <ronnie_> Dan: Annotate our use cases based on which involve controlling the content. Indexers do. Book club, students, reading system don't.
15:33:27 <ronnie_> ... If there are various efforts to solve the problem and we have a partial solution through CFI and make partial progress and defer the rest till there's a clearer answer.
15:34:16 <ronnie_> ... CFI has broad usage but not reader-friendly, maybe we could make them more friendly with a dictionary. Let Chrome figure out their anchor thing separately.
15:34:32 <ronnie_> Wendy: I like this idea.
15:35:29 <ronnie_> Dan: This feels like offline work on a spreadsheet. The tool at the end of the chain gives the affordance. Indexers are much more at the beginning.
15:35:39 <ivan> q+
15:35:46 <pilarw2000> q+
15:35:48 <wendyreid> ack ivan
15:35:53 <ronnie_> ... Locator provided by reading system or by author, another dimension.
15:36:33 <wendyreid> ack pilarw2000
15:36:36 <dlazin> We are basically making a table like this: https://en.wikipedia.org/wiki/Comparison_of_e-readers#Electronic-paper_displays
15:36:37 <wendyreid> ack pilarw
15:36:38 <ronnie_> Ivan: The stability of the pointers is also another dimension. Books sometimes change slightly. CFI is very bad at that. Whether a tool is stable against change or not is important.
15:37:15 <ronnie_> Pilar: Increasingly ePub texts don't have enough anchors. If you have section numbers or paragraph numbers, but many don't.
15:37:39 <ronnie_> Ivan: Technically speaking, that's not enough. If you just have text, you still need anchors for identifiers.
15:38:17 <ronnie_> Pilar: Point taken. Do we have an additional use case, that has section or paragraph numbers but just needs the linking put in? That's different for text without any.
15:38:40 <ronnie_> Ivan: The content is in HTML. The real problem is the missing anchors, not the missing structure.
15:38:56 <ronnie_> Pilar: In an index you need the anchor and what you'll call it for the anchors.
15:39:24 <ronnie_> Wendy: Do we create an algorithm that numbers the <p> tags. But then you need to label it.
15:39:40 <ronnie_> Dan: In a print book index, the label is a page number.
15:39:54 <ronnie_> Pilar: or line numbers in poetry or section for legal texgt.
15:40:33 <ronnie_> Dan: As discussed on page x, y, z - they don't mean anything in ePub. They work nicely as labels for anchors in text.
15:40:59 <ronnie_> Brady + Dan: What three words is a fine solution. Order numbers are done like that.
15:41:13 <ronnie_> Wendy: Netlify assigns domains based on that as well.
15:41:20 <ronnie_> Brady: Not very international
15:41:48 <ronnie_> Dan: Multiple dictionaries could resolve to the same CFI thing.
15:43:24 <ronnie_> Brady: Responding to Ivan. CFI is fairly stable across changes. The path could include all the IDs, text as well. As long as you leave the paragraph IDs in place, it's OK with modifications for the content. It breaks only if you change the IDs. The horrible path could include all the interim IDs.
15:44:13 <ronnie_> Ivan: In the CFI itself you encode the path in the doc. If I add a new <p> in the middle, all the CFIs pointing to a later paragraph will be wrong.
15:44:50 <ronnie_> Brady: Worst case if there are no IDs, but even then you could include text from the target, you could look for it in a neighbouring paragraph.
15:45:26 <ronnie_> Ivan: That's a runtime implementation. It's not in the spec. There's a way to implemetn a redirection, but the original identifier is incorrect in a way.
15:45:43 <ronnie_> ... That means if we did a note, we'd have to describe the processing.
15:46:17 <ronnie_> Brady: The processing model was considered too risky to include, but there should be one.
15:46:32 <ronnie_> Ivan: What's the next step?
15:46:45 <pilarw2000> q+
15:46:49 <dlazin> q+
15:47:12 <ronnie_> Wendy: Identify which use cases intersect at which specific points: control or lack of control; creates and uses the locators.
15:47:22 <wendyreid> ack pilarw
15:47:32 <ronnie_> ... Google Doc isn't good for this, a sheet will work.
15:48:17 <ronnie_> Pilar: If an indexer wants to use the CFIs as targets, how to label them in the index?
15:48:28 <ronnie_> ... CFIs don't show up to human reader
15:48:46 <ronnie_> Ivan: A URL isn't there for the human reader either.
15:49:08 <ronnie_> Pilar: If the indexer wants to use the embedded index tags, how should they be labelled?
15:49:31 <duga> q+
15:49:31 <ronnie_> Wendy: Pilar will add this one.
15:49:32 <wendyreid> ack dlazin
15:49:59 <duga> q-
15:50:13 <ronnie_> Dan: We shouldn't prejudge this. But from responsiveness to the original feature request and ePub spec, we won't solve page numbers.
15:50:25 <ronnie_> ... We might have a better web focus.
15:50:54 <ronnie_> ... We might have a better web solution. We need a way to refer to "page" locators.
15:51:31 <ronnie_> Pilar: Indexers were largely fine with using the paragraph numbers in production, but how long will they last?
15:51:52 <ronnie_> Ivan: I have a script that turns W3 recommendations into ePub. No page numbers.
15:52:20 <ronnie_> Wendy: w3C recommendations are better prepared, because of section numbering. The audiobook spec can refer to subsections.
15:53:33 <ronnie_> ... Two things we want to figure out: 1) how do we make these references? visually to the user, the author, the indexer? 2) also a visual representation of the locators semantically.
15:54:20 <ronnie_> Wendy: Breaking up the use cases by control. Also, what could we take from CFI to make it a parsable property?
15:54:38 <ronnie_> Ivan: You don't want an average human to see a URL and parse that either.
15:54:50 <ronnie_> Wendy: I know a URL means this site, this page
15:55:18 <ronnie_> Brady: Sortable feature of CFI. Any label that I can look at and see if it's later is helpful.
15:55:44 <ronnie_> Pilar: Sequential and also size. The equivalent of a page range for a large section or a small subsection.
15:55:54 <ronnie_> Ivan: CFI won't get that, only the sort.
15:56:04 <ronnie_> Pilar: Label show a beginning and an ending?
15:56:11 <wendyreid> https://docs.google.com/spreadsheets/d/1KO-HyLGUUw36F-ruAARHNiPO1aUJCCNeTv3zxGtjuHw/edit?usp=sharing
15:56:18 <ronnie_> Wendy: Page numbers are semantically rich
15:56:46 <ronnie_> ... I created a spreadsheet where we can resort the use cases.
15:57:03 <ronnie_> Pilar: I'll add my use case to the Google Doc.
15:57:34 <ronnie_> Wendy: We need to make CFI more human readable or to have a more human readable equivalent.
15:58:02 <ivan> rrsagent, draft minutes
15:58:02 <RRSAgent> I have made the request to generate https://www.w3.org/2021/04/21-epub-locators-minutes.html ivan
16:00:55 <ronnie_> Dan: Presentation is an issue. We have to thinking about how books work. I was imagining an index where each thing is collapsed by default, you'd see a window to the page that shows the options.
16:01:20 <CharlesL> CharlesL has left #epub-locators
16:01:32 <ronnie_> Pilar: Dan, what you described was in the last ePub spec for indexes.
16:01:43 <ivan> rrsagent, draft minutes
16:01:43 <RRSAgent> I have made the request to generate https://www.w3.org/2021/04/21-epub-locators-minutes.html ivan
16:03:43 <ronnie_> At 11:59 Ronnie: Added the question of presentation. I did an index of a large portal of information, and you used typeahead to find the index entry you wanted, and when you clicked the locator, you got the specific hit.
18:25:40 <Zakim> Zakim has left #epub-locators