Meeting minutes
w3c/epub-specs#3071
Wendy Reid: this is "offer a human readable way to identify the annotation"
… we talked about this a few weeks ago, about providing breadcrumbs and hierarchy to find the target of the annotation
Laurent Le MeurM: proposed adding two new pieces of metadata to help people find the target
Laurent Le MeurM: we have two solutions there, as strings, one which will be different from one rs to another
… the second one will be more structured and is a little harder to define. Should it be a page? or a locator?
… I recommend we keep a single label and let the reading system make the magic
Hadrien Gardeur: What is this for? If we're talking about display in an RS, it probably isn't needed. The main use would be when you don't have access to the book
… in that case, some kind of label which may not always be useful, progression in the book might be the best we can do. I don't recommend using page or page list because they are not reliable
Wendy Reid: I agree progression is challenging, and page list isn't reliable. I know there are a lot of books without page lists and this is a challenge. Maybe percentage plus the headings would be sufficient
… in the context of a reading system it might not be needed, but when it is exported to another tool, or the exported results themselves, when you look at annotations as an export, you need context to know what you are looking at. It's not adequate to see just the note
Brady Duga: I understand why we might want these, page in the current proposal is only for fixed layout, which might be not annotated very often. We should figure out an answer, we might not want to create new strings. We have had problems with strings in the package before by allowing raw strings with no markup. Internationalization will make this more
… complex. If we're starting to introduce new strings that will be displayed with no markup, it may not be properly represented
Hadrien Gardeur: I hear you Brady, I feel like there is not much we can do about this. Whenever we try to use HTML it is useless, and usually converted back to a string. We need to find a way to do this with unicode but that isn't our problem to solve. We keep going back to this, but I think it is a unicode problem and not ours.
Wendy Reid: where does that leave us?
Brady Duga: we put this on the agenda because it didn't get much traction. And that is happening here. Do we have someone who really champions this? Is it really needed?
… do we want to throw it out there and see if it works?
Laurent Le Meur Le Meur: As we have seen I think this feature ins't about the interchange between reading systems, more about exporting them to something that is human readable. We know that human readable export from a reading system may be something like HTML and used for things like citations.
… since we have decided not to conflate annotations with citations maybe we don't need this
Wendy Reid: Just exporting annotations essential raw, and import it into HTML or Markdown, all I will get is which book its attached to, a locator that isn't human readable, and I'll know if it is a highlight or a note. That is not enough to be usable. We see that in how platforms current export annotations. We need at least some human readable
… parts. Kobo mentions the chapter and that gives me a chance to find the target
Hadrien Gardeur: This has to do with whether you can present this outside a reading system or not? Do we want the format to be merely interchange or do we want it to be useful outside RS? If this is just about exchange we don't need any of this. What's the goal? When we answer that, it becomes easier to decide.
Laurent Le Meur: When we started designing the export mechanism, I pictured it as a source for creating a human readable output in PDF or HTML. I realize that the reading system is capable of doing this. Kobo has multiple file formats on export, including ones that are human readable. We could leave this to the reading system to decide.
Wendy Reid: Hadrien Gardeur has asked an important question. What is the primary use case we are targeting. They are not mutually exclusive, but what are we leaning into more? We want a usable format, there has been a request for this for a long time. What are their goals? Exporting or interoperability?
Susan Neuhaus: I worry a little about leaving export to reading systems, if our goal is interchangability, reading systems might be tempted not to export things in a human readable way
… or not to other systems, just to try to maintain their business model
… if we leave it to the reading system we may not get what we want to get
Brady Duga: the suggestion isn't to leave it up to the reading system, but to leave the options up to the user. You can choose HTML, PDF or something like .annotation(machine readable)
… we need to think about the people who read books, if we are targeting regular people we have lost everyone with JSON LD. Most people will look for a program to open it with. If we want a human readable format, we need to go back to basics. We need a clear separation between human and machine readable
Hadrien Gardeur: we won't get human and machine readable in the same format. The bigger question is "Can I generate something human readable from the machine readable file"? A reading system has all it take to generate either.
… however if I want to aggregate all my bookmarks and notes somewhere else, how can I do that. It is difficult to answer what users want, it is hard to answer because we don't even have a machine readable file now. We are just guessing if this is needed.
George Kerscher: I'm concerned with complexity for the end user. I know students want to be able to save these for study purposes, and it can be done to day with different programs. If I want to do something different, like moving annotations to another reading system, that's more complicated now. But it has to be human readable
Laurent Le Meur Le Meur: We guarantee that with our format that you can transfer your annotations from one RS to another as long as you have the same book in each. What we do not warranty is that without having the book that you will be able to make meaningful use of the annotations. You will have not context for what the annotation is about. Can this format be
… changed into something human readable without the help of another application
Wendy Reid: Do we have something in the annotations system that would allow us to take .annotations file and make it into something human readable. There will always be an app involved. Do we have the metadata that will make that possible. We can't allow for all publications, but we should be able to make this possible. The question is is this
… metadata valuable enough to be in the package
Laurent Le Meur Le Meur: If I read in Thorium, and I have the same book in Zotero, and I export my annotations from Thorium, now I see my book and notations in Zotero, now I can export both the annotations and the citations.
Hadrien Gardeur: If I am reading a book this DRM, and it expires, and I try to bring it in to another system with the same book, I may not be able to use the annotations. If I don't have access to the file, I may not be able to make a human readable version
George Kerscher: If I make a set of annotations and the reading system gives me the option to exchange with other RS or save to MSW for example, that is doable? That ability to get the human readable piece out of it would be essential
Brady Duga: does the current spec have the text of the book in the current annotation"
Laurent Le Meur Le Meur: Not necessarily, it depends on the selector
Brady Duga: that is text before and after, right?
Laurent Le Meur Le Meur: yes, but we have a caution about protected text
Brady Duga: if I make a blue highlight on page 3 of chapter 2, I can know that there is blue text, but without the text it is useless. If we want to do this, we will have to include the text. Or the human readable output would look kind of silly
Charles LaPierre: Couldn't a person annotate a whole chapter, export it and the publisher object?
Brady Duga: we don't discuss DRM here, that is between the Reading System and their DRM provider. It means that certain books you cannot reconstruct your annotations because of choices made by the rights holder
Wendy Reid: There are systems where there is a limit on exporting content. You can select it, but all you get is something like 600 characters. Based on an agreement between the RS and the publisher.
Laurent Le Meur Le Meur: We will not create something that makes protecting books with a DRM impossible. We don't talk about copyright protection, but we don't do anything to make it impossible
Hadrien Gardeur: You're saying that in most reading systems that I can export 600 characters at a time? If we can do that over and over again we could answer Brady's question
… 600 is enough to find an annotation target
Brady Duga: I know systems that have a quota for the percentage of the book and every time you copy or export from a DRM book. It could be painful in this situation, if you mark up a lot. It doesn't come up in non fiction. I suspect the limit is around 20% Most systems won't let you select more than one page at a time. There are easier ways to steal
… content.
Hadrien Gardeur: 20% is often the limit for previews, If I have the 50 first and last characters and an elipses between or whatever, that is already enough. It is doable for something that is human readable if we use the right selectors. It is nice to have the headers, but the text is more important.
… we still haven't answered the initial question, do we want to make something human readable?
Laurent Le Meur Le Meur: Use case 6, generating citations, either we remove this use case because a User can export this easily from the reading system or we implement it.
Wendy Reid: this is a philosophical question, what do we want from annotations?
Laurent Le Meur Le Meur: I propose that we focus on exchange for now, and leave it open to future versions
Hadrien Gardeur: Is interchange the most important use case? We know people will use human readable export because they already do it
Laurent Le Meur Le Meur: If this fails, if no one uses it, it won't be because we make human readable content also
Hadrien Gardeur: I disagree, if my main use is Zotero, and I can't do that, then it is useless to me
Gautier Chomel_: if we don't include this we might end up with something messy
Laurent Le Meur Le Meur: If we remove this will will simply have to change our specs
Abe Jellinek: people ask for human readable output all the time, we've added some support for other readers, its not ideal as we are parsing their native formats. One of the advantages of Zotero is that the annotations are separate from the books on our server and are portable.
Wendy Reid: is display part of the interchange?
Abe Jellinek: Yes, but not as often
… what people have is annotations on another RS and they want them on Zotero, we will have them as machine readable formats, and display is another issue
Hadrien Gardeur: You need information to make the display. I ask what is the point of doing something if I can't get anything out of it? When you aggregate something from Kindle do you get enough information to show the user a list and provide something human readable? Why?
… what makes an annotation format usable enough that you can show them the annotation and make it human readable?
Abe Jellinek: the bare minimum is a text snippet, we get into issues if we expect that to be all. We use whatever the RS considers its native locator. What Kindle generates right now is almost unusable, last I checked its a txt file with page numbers, quotes and comments in human readable format. KOReader uses a weird take on XPath for CFI, Calibre uses compliant CFI and it makes import easier
Wendy Reid: Abe Jellinek that was incredible helpful thank you!