Re: [External] My opinions about the future of EPUB 3.2

I think Laurent¹s idea of progressive levels of ³compliance² makes sense,
but the gotcha is that it will require a VERY large effort to implement the
tests and THEN to run them, log the results and somehow disseminate those
results.

Laurent is also correct that EPUBCheck is about compliance, not interop ­
though it effectively does some interop checking.  The larger problem is
that the Test Suite (TS) is not nearly complete enough.  It lacks three key
elements:
* There are many many aspects of EPUB that aren¹t checked simply because
EPUBCheck is way out of date.
* The TS tests all the features in isolation.  This is fine as a first step,
but in practice (based on Readium¹s experience) the vast majority of
problems are due to two or more features interacting.  I can¹t remember the
last time we found a bug in Readium because we ran the Test Suite (which we
do frequently).
* There are essentially no negative tests, I.e. tests which test whether an
invalid or broken test  is correctly handled.
Now, you could argue that even an improved expanded is about compliance, not
interop, but IMHO, one needs such tests to verify that a single EPUB
document, once created, renders correctly on a wide variety reading systems.

Ric


From:  Laurent Le Meur <laurent.lemeur@edrlab.org>
Date:  Tuesday, October 30, 2018 at 9:04 AM
To:  W3C Publishing Business Group <public-publishingbg@w3.org>
Cc:  Jeff Jaffe <jeff@w3.org>
Subject:  Re: [External] My opinions about the future of EPUB 3.2
Resent-From:  <public-publishingbg@w3.org>
Resent-Date:  Tue, 30 Oct 2018 16:04:31 +0000

I don't agree that EPUBCheck is the proper tool for checking the
interoperability we are looking for.

EPUBCheck is a validation tool, not an interop checker. This tool is there
for a long time, EPUB files comply to its verdict, and still authors and
publishers complain that they must test on several reading system to avoid
very bad surprises.

-> Interoperability must be checked at the level of reading systems, not at
the level of the EPUB files.
 
Note: if we define a feature interoperable when 2 reading systems implement
it correctly, authors and publishers will still complain, and will be right
to do so. We have to be more ambitious about the level of interop the
industry need, if this is possible.

Is is really impossible to create compliance levels for reading systems,
with a basic level applicable to any reader (with e.g. reflow only, images
but no audio/video, pagination, support of links), plus a set of
progressively more complex levels, each level getting a small set of
features, features which the CG will have listed and described in the first
phase of its work? I have already introduced this idea without getting any
positive feedback, but no better alternative solution has been provided so
far. 

Laurent Le Meur
EDRLab

> Le 30 oct. 2018 à 16:19, Jeff Jaffe <jeff@w3.org> a écrit :
> 
> 
> 
> On 10/29/2018 2:47 PM, Johnson, Rick wrote:
>> Jeff,
>>  
>> I believe that defining interoperability of the spec (on a feature by feature
>> level) needs a more limited focus, which may differ from how the W3C has
>> approached interoperability in the past.  What I am advocating for is
>> interoperability of the spec for creation and distribution of a file, and to
>> stop there (obviously highly dependent on EPUBcheck for this!).  Is it
>> possible within the current process document (and common practice) to stop
>> there, and avoid the discussion of Œinteroperability within the reading
>> systems¹ as a requirement?
> 
> I'm not sure I fully understand the question - but if I do - the answer is
> Yes.
> 
> W3C Recommendations specify some behavior, and to get to a REC level requires
> that the spec has been implemented interoperably.  Typically, we would like
> for test cases which validate at a feature by feature level that the spec has
> been successfully implemented.  I believe this is what you are referring to
> when you say interop for creation and distribution.
> 
> I'm not sure what you mean by "interoperability within the reading systems" -
> but I am interpreting that to be a sort of compliance and certification test
> whereby a standards organization certifies that an entire implementation (in
> this case a reading system) has implemented an entire spec.  That is generally
> beyond the scope of W3C Recommendations.
> 
>>  
>> -Rick
>>  
>>  
>> From: Jeff Jaffe <jeff@w3.org> <mailto:jeff@w3.org>
>> Date: Monday, October 29, 2018 at 11:15 AM
>> To: Laurent Le Meur <laurent.lemeur@edrlab.org>
>> <mailto:laurent.lemeur@edrlab.org> , Ric Wright <rkwright@geofx.com>
>> <mailto:rkwright@geofx.com>
>> Cc: Brian O'Leary <brian@bisg.org> <mailto:brian@bisg.org> , W3C Publishing
>> Business Group <public-publishingbg@w3.org>
>> <mailto:public-publishingbg@w3.org> , "Johnson, Rick"
>> <Rick.Johnson@vitalsource.com> <mailto:Rick.Johnson@vitalsource.com> , Ivan
>> Herman <ivan@w3.org> <mailto:ivan@w3.org>
>> Subject: Re: [External] My opinions about the future of EPUB 3.2
>> Resent-From: <public-publishingbg@w3.org> <mailto:public-publishingbg@w3.org>
>> Resent-Date: Monday, October 29, 2018 at 11:14 AM
>>  
>>  
>> 
>>  
>> On 10/29/2018 1:30 PM, Laurent Le Meur wrote:
>>> Whatever the business situation has been or will be, I learned from Jeff at
>>> the TPAC that to become a standard, each individual feature must be
>>> implemented twice, which could be a practical definition of
>>> interoperability. Hadrien reminded me that today.
>>>  
>>> But I don't find this exact definition in the new version of the W3C Process
>>> Document (Implementation experience).
>>> https://www.w3.org/2018/Process-20180201/#implementation-experience
>>>  
>> 
>> Indeed, the process document is written in a way to give the Director
>> flexibility.  We want to make sure that there is sufficient implementation
>> experience - but each situation is different - so instead of having strict
>> requirements, we list considerations that the Working Group and Director
>> should think about.
>> 
>> The second consideration of Section 6.2.4 (which is neither necessary nor
>> sufficient by itself), is the question whether there are independent
>> interoperable implementations.  Generally, the Director is looking for
>> positive answers to these questions!  If the answer to this consideration is
>> - Yes - then the yields the oft-quoted two interoperable implementations.  If
>> the answer is - No - a specification could still advance, but has a higher
>> hill to climb.
>> 
>> 
>>> Cordialement, 
>>>  
>>> Laurent Le Meur
>>> EDRLab
>>> 
>>> 
>>>> Le 29 oct. 2018 à 17:49, Ric Wright <rkwright@geofx.com> a écrit :
>>>>  
>>>> Sorry, my experience is that the booksellers will fight side-loading tooth
>>>> and nail.  Your third paragraph is exactly what we tried to achieve at
>>>> Adobe and it was, for all intents and purposes, a failure. The booksellers
>>>> did not want it.  Perhaps it will be different this time around, but I am
>>>> skeptical.
>>>>  
>>>> Ric
>>>>  
>>>>  
>>>> From: Laurent Le Meur <laurent.lemeur@edrlab.org>
>>>> Date: Monday, October 29, 2018 at 9:37 AM
>>>> To: Brian O'Leary <brian@bisg.org>, W3C Publishing Business Group
>>>> <public-publishingbg@w3.org>
>>>> Cc: "Johnson, Rick" <Rick.Johnson@vitalsource.com>, Ric Wright
>>>> <rkwright@geofx.com>, Ivan Herman <ivan@w3.org>
>>>> Subject: Re: [External] My opinions about the future of EPUB 3.2
>>>> Resent-From: <public-publishingbg@w3.org>
>>>> Resent-Date: Mon, 29 Oct 2018 16:38:04 +0000
>>>>  
>>>> All reading systems independent from booksellers provide side-loading and
>>>> therefore promote interoperability within the marketplace. Look at Bluefire
>>>> reader, Lis-a reader, Bookari, Aldiko, FolioReader, FBReader, Bookeen, TEA
>>>> or Tolino e-readers etc.
>>>>  
>>>> And some reading systems published by booksellers are also open to this
>>>> (LEA Reader from Adilibre/Albin Michel comes to my head but they are
>>>> certainly many others).
>>>>  
>>>> For a customer, being able to select one favorite app for its whole
>>>> personal library is a great feature; and for booksellers, making so that
>>>> customers choose their app (with a direct access to their catalog) should
>>>> be a goal. 
>>>>  
>>>> Cordialement, 
>>>>  
>>>> Laurent Le Meur
>>>> EDRLab
>>>> 
>>>> 
>>>>> Le 29 oct. 2018 à 17:11, Brian O'Leary <brian@bisg.org> a écrit :
>>>>>  
>>>>> If interoperability within the marketplace is never going to happen, I
>>>>> guess we can close up shop now.
>>>>>  
>>>>> On Mon, Oct 29, 2018 at 12:07 PM Johnson, Rick
>>>>> <Rick.Johnson@vitalsource.com> wrote:
>>>>>> On this point, it would be helpful to understand the definition (and
>>>>>> example use cases) for interoperability here.  There are two very
>>>>>> different possibilities in my mind (and probably more!):
>>>>>> 1. Interoperability before reaching the marketplace: A content creator
>>>>>> can deliver a file to any reading system for their ingestion, and
>>>>>> Œinteroperability¹ will allow them to create just one file for
>>>>>> distribution.
>>>>>> 2. Interoperability within the marketplace: A reading system can open
>>>>>> content being distributed in the marketplace from another reading system.
>>>>>>  
>>>>>> The first is a logical goal.  The second is never going to happen!
>>>>>>  
>>>>>> Are there other options?
>>>>>>  
>>>>>> -Rick
>>>>>>  
>>>>>> From: Ric Wright <rkwright@geofx.com>
>>>>>> Date: Monday, October 29, 2018 at 8:57 AM
>>>>>> To: Ivan Herman <ivan@w3.org>, Laurent Le Meur
>>>>>> <laurent.lemeur@edrlab.org>
>>>>>> Cc: W3C Publishing Business Group <public-publishingbg@w3.org>
>>>>>> Subject: [External] Re: My opinions about the future of EPUB 3.2
>>>>>> Resent-From: <public-publishingbg@w3.org>
>>>>>> Resent-Date: Monday, October 29, 2018 at 8:56 AM
>>>>>>  
>>>>>> These are all good points, but I guess I wonder what is driving the
>>>>>> desire for interoperability?  Or perhaps more accurately, IS there any
>>>>>> desire for interoperability?  As some of you are aware, I led the effort
>>>>>> at Adobe with ADE, ACS4 and RMSDK.  One of the huge takeaways was that
>>>>>> the customers for RMSDK and ACS positively didn¹t WANT interoperability.
>>>>>> They all wanted to be in separate silos.  We had to fight hard to make
>>>>>> them support side-loading, for example.
>>>>>>  
>>>>>> This is a slightly different aspect of ³interoperability² since Adobe
>>>>>> ensured that the underlying reading system (RMSDK/ACS) was fully
>>>>>> interoperable.  But even then the vendors all wanted to create walls
>>>>>> around their silo.
>>>>>>  
>>>>>> My point is that achieving interoperability will require some impetus for
>>>>>> reading systems and publishers to ensure their solutions are
>>>>>> interoperable.   What is that impetus?  Why will reading systems and
>>>>>> publishers spend non-trivial time, money and opportunity to achieve
>>>>>> interoperability?  If we can¹t answer that then it seems like we are dead
>>>>>> in the water.
>>>>>>  
>>>>>> Ric
>>>>>>  
>>>>>> From: Ivan Herman <ivan@w3.org>
>>>>>> Date: Sunday, October 28, 2018 at 10:45 PM
>>>>>> To: Laurent Le Meur <laurent.lemeur@edrlab.org>
>>>>>> Cc: W3C Publishing Business Group <public-publishingbg@w3.org>
>>>>>> Subject: Re: My opinions about the future of EPUB 3.2
>>>>>> Resent-From: <public-publishingbg@w3.org>
>>>>>> Resent-Date: Mon, 29 Oct 2018 05:45:39 +0000
>>>>>>  
>>>>>> Laurent,
>>>>>>  
>>>>>> I do not disagree with your analysis. Yes, the interoperability issue is
>>>>>> the main challenge, regardless of where it is handled.
>>>>>>  
>>>>>> However, if the decision is that, at some point, the existence of a W3C
>>>>>> Rec is required, there is no problem doing things in parallel, namely
>>>>>>  
>>>>>> 1. the CG can work right now in defining the interoperability
>>>>>> requirements, creating tests, etc
>>>>>> 2. we (as a collective 'we' at this point) can start hammering out the
>>>>>> details of what a Rec track work would mean, write a charter that
>>>>>> provides the necessary and good arguments on why EPUB 3.2 should be a Rec
>>>>>> track, how that would modify the direction of the current WG, etc. And,
>>>>>> of course, the new charter should be accepted by the W3M management and
>>>>>> the AC, which rarely goes without further discussion to find the right
>>>>>> consensus.
>>>>>>  
>>>>>> Step (2) will require lots of time and energy. *If* we agree on the goal,
>>>>>> there is no reason to wait for (1) to complete. Whatever the CG does in
>>>>>> (1) can be input to a new WG for further work right away.
>>>>>>  
>>>>>> cheers
>>>>>>  
>>>>>> Ivan
>>>>>>  
>>>>>> P.S. Let me also add one more thing to your 2/ below. A Rec version of
>>>>>> EPUB3.2 is not _only_ of interest in view of ISO. I think the Patent
>>>>>> Policy protection attached to a Rec may be significant in future for
>>>>>> implementors, as well as the guarantee to provide a public access to the
>>>>>> specifications once and for all (which may be a hurdle for an ISO
>>>>>> document).
>>>>>>  
>>>>>>  
>>>>>>  
>>>>>>> 
>>>>>>> On 24 Oct 2018, at 11:22, Laurent Le Meur <laurent.lemeur@edrlab.org>
>>>>>>> wrote:
>>>>>>>  
>>>>>>> Hi, 
>>>>>>>  
>>>>>>> To use the new WG motto: what is the problem we're trying to solve?
>>>>>>>  
>>>>>>> I see two problems to solve here:
>>>>>>> 1/ raise the level of interoperability of EPUB 3 reading systems
>>>>>>> 2/ get EPUB 3.2 standardized by ISO, in order to get Asian adhesion to
>>>>>>> this version of the standard.
>>>>>>>  
>>>>>>> Getting EPUB 3.2 standardized by the W3C is therefore not a solution to
>>>>>>> the problems; the fact is that resolving the first problem is required
>>>>>>> to get EPUB 3.2 as a rec, and getting EPUB 3.2 as a rec is a possible
>>>>>>> step in the resolution of the second one.
>>>>>>>  
>>>>>>> Therefore I propose to first tackle the first problem,
>>>>>>> theinteroperability issue.
>>>>>>>  
>>>>>>> For this purpose we need IMO to
>>>>>>> a/ define what interoperability really means in our case; this is not so
>>>>>>> obvious, as there is a large diversity of reading systems (especially
>>>>>>> those based on browser engines, those with custom rendering engines,
>>>>>>> those with no visual engine (audio or braille UAs); we may have to
>>>>>>> defines different classes of reading systems and different
>>>>>>> interoperability levels.
>>>>>>> b/ list every feature defined in EPUB 3.2, with their testing
>>>>>>> requirements. Be careful about what we want to test: do we (still) want
>>>>>>> to replicate the html/css "can I use"? for custom rendering engines,
>>>>>>> which implement a subset of CSS (and maybe HTML5), this could be
>>>>>>> necessary ...
>>>>>>> c/ modify the existing (create a new) EPUB test suite, a set of EPUB
>>>>>>> samples which will help testing each features individually
>>>>>>> d/ update the epubtest.org <http://epubtest.org/>  service to handle the
>>>>>>> new test suite.
>>>>>>>  
>>>>>>> The Publishing CG seems to me the proper place for this work, as it has
>>>>>>> released EPUB 3.2 (and therefore is now free), can get help from
>>>>>>> everybody in the industry (BISG ...), and it does not require
>>>>>>> rechartering the WG.
>>>>>>>  
>>>>>>> It would be of tremendous help for the industry to get it done.
>>>>>>>  
>>>>>>> In parallel, we should try to assert the difficulty to get EPUB 3.2
>>>>>>> directly prepared for ISO standarization (in men/hours).
>>>>>>>  
>>>>>>> Then and only then, after the main reading systems on the market have
>>>>>>> been tested against the new test suite, we will be able to assert the
>>>>>>> difficulty to get EPUB 3.2 standardized by the W3C (in men/hours), i.e.
>>>>>>> if there are two conforming implementations for each feature of the
>>>>>>> standard. This work can be made by the WG.
>>>>>>>  
>>>>>>> With this data, we'll be able to decide if direct ISO standarization is
>>>>>>> harder or simpler than W3C rec + W3C to ISO standardization.
>>>>>>>  
>>>>>>> During this period, the WG will be able to focus on an urgency for the
>>>>>>> industry: "WP and EPUB for audiobooks", which could even be released in
>>>>>>> 2019 maybe (because the spec is simpler that "WP and EPUB for any type
>>>>>>> of ebook"). And we may find the time to start working on "WP and EPUB
>>>>>>> for comicbooks" also. Without rechartering...
>>>>>>>  
>>>>>>> Cordialement,
>>>>>>>  
>>>>>>> Laurent Le Meur
>>>>>>> EDRLab
>>>>>>>  
>>>>>>> 
>>>>>>> Le 24 oct. 2018 à 02:11, MURATA Makoto <eb2m-mrt@asahi-net.or.jp> a
>>>>>>> écrit :
>>>>>>>  
>>>>>>> Folks,
>>>>>>>  
>>>>>>> I read the draft minutes of the joint F2F of the publishing working
>>>>>>> group 
>>>>>>> and the publishing business group with interest.  (Ivan, special thanks
>>>>>>> to your timely draft minutes! )  Here are my opinions about the future
>>>>>>> of 3.2.
>>>>>>> ...
>>>>>>>  
>>>>>>  
>>>>>> 
>>>>>> ----
>>>>>> Ivan Herman, W3C
>>>>>> Publishing@W3C Technical Lead
>>>>>> Home: http://www.w3.org/People/Ivan/
>>>>>> mobile: +31-641044153
>>>>>> ORCID ID: https://orcid.org/0000-0003-0782-2704
>>>>>>  
>>>>> 
>>>>>  
>>>>> --
>>>>> Brian F. O'Leary
>>>>> Executive Director, Book Industry Study Group
>>>>> 232 Madison Avenue, Suite 1400
>>>>> New York, NY 10016
>>>>>  
>>>>> (646) 336-7141 office
>>>>> (973) 985-9880 mobile
>>>>  
>>>  
>> 
>> 
> 
>