Re: What do Web Publication User Agents Do? How Do We Test Them?

Hi Dave,

One of the issues I have with this simple requirement is that non-linear web publications are already a big big thing on the web: wikis, API and technical docs, and many educational resources.

These aren’t affected by the requirement to provide a default reading order as they’d just have a reading order of one (the entry point) and then provide their own navigation while still being able to take advantage of other things the WG is hoping will be available (like customisability, annotation, and offline).

Of course, none of that is a problem if, as I’d prefer it, most publication features aren’t standardised as publication-specific features but web features that websites can opt in and use if needed. That makes them easier to polyfill individually and makes campaigning browser vendors for implementation easier as they can pick and choose. Which in turn would make publication-specific UAs easier as the higher likelihood of individual features being entirely taken care of by the browser makes their implementation simpler.

Generally, I favour Hadrien’s overall take on this in a later email in this thread. Specifically the fact that publications as outlined by the requirements have a bunch of unsolved problems that we can’t really solve in the near term. And we shouldn’t try to solve them. At least not yet. Instead we should:

* Focus on defining an absolutely minimal manifest and metadata. 
* Break everything up into a list of individual requirements and tackle them one by one, each in the most appropriate way. 
* Preferably specify each in a way that allows it to be implemented independently of other requirements, purely as a practical matter.

The implication of this is that pretty much everything about web publications are, from the web platform perspective, CANs or SHOULDs at best. If you model them as granular features, the idea of making hard requirements of the pre-existing web development community makes little sense.

- best
- Baldur Bjarnason
  baldur@rebus.foundation


> On 13 Aug 2017, at 22:47, Dave Cramer <dauwhe@gmail.com> wrote:
> 
> Hi Everyone,
> 
> The core of the idea of a web publication was expressed in requirement 7 (http://w3c.github.io/dpub-pwp-ucr/index.html#r_single) of our use case document: 
> 
> “User agents must treat a Web Publication as a single logical resource with its own URL, beyond the references to individual, constituent resources.”
> 
> But what does this mean in practice? 
> 
> We've talked about the abstract manifest providing a list of primary resources and their default ordering.  But how would you test such an assertion? What is a user agent supposed to do with such information? This also gets to the question of why we need web publications at all. What do we need that the web doesn't already do?
> 
> The UCR document listed several aspects of this:
> 
> 1. The scope of search should be the entire publication.
> 
> 2. Personalization choices should apply to the whole publication (Personalization is requirement 11)
> 
> 3. CSS counters should operate across the entire publication.
> 
> 4. Assistive technologies should treat a publication as a single unit.
> 
> 
> I would propose that there is another fundamental, and very simple, requirement: can you access all the primary resource content without clicking links?  
> 
> This sounds crazy in the context of the web. But this is what we have from every ebook reading system—"turn the page" or press the "next" button, and you can go through the entire contents, screen by screen. No hunting for blue underlined text. No going back to a table of contents, figuring out what the next chapter is, and clicking on a target that most likely occupies less than 1% of the screen area. 
> 
> I've mentioned earlier that I think Jeremy Keith's "Resilient Web Design" is a great example of a book on the web today. But you have to click links to get from chapter to chapter, and this is what makes it different than today's ebooks. (And as I've noted before, you can get through the book in Opera 12 without clicking links, due to UI around rel=prev/next). 
> 
> I would add a few more testable assertions about a web publication:
> 
> 5. A web publication user agent should remember where the user is, and restore that state any time a user navigates back to the WP.
> 
> 6. The table of contents should be available from every primary resource. (Requirement 13)
> 
> 5. A web publication should have a shareable URL.
> 
> 6. A web publication should be readable while offline. (Requirement 6)
> 
> 7. A web publication should allow annotations, including highlights, notes, and bookmarks.
> 
> What else are we missing? 
> 
> Dave

Received on Monday, 14 August 2017 16:19:54 UTC