cross links randomly become footnotes

[AlanHK]

I was crosslinking terms in a glossary.
Editing an epub2 in Sigil. Everything worked as expected in Sigil and viewing the epub in calibre's reader.

Export to AZW3 with Kindlegen, view on PW-3, and most of the linked words are links that take me to that term, a few instead pop up footnotes.

Cannot see any difference between the code for ids or links.

Example: 3 entries below.
Viewed on Kindle, in boson the link for "fermion" is a footnote. The link for "Kozuch Theory" is a jump link.

<p class="gloss"><a id="boson"></a><b>boson</b>. All elementary particles can be classified as either bosons or <a href="#fermion">fermions</a>; the bosons include photons and gluons. The quantum wave function for two or more identical bosons is unchanged if any two particles are swapped, and the wave function for a single boson is unchanged if the particle is rotated by 360 degrees. Bosons have a spin which is an integer multiple of the fundamental unit of angular momentum. In <a href="#KozuchTheory">Kozuch Theory</a>, all these properties arise from the topology of the particle’s wormhole.</p>


<p class="gloss"><a id="fermion"></a><b>fermion</b>. All elementary particles can be classified as either <a href="#boson">bosons</a> or fermions; the fermions include electrons and quarks, and composites of three quarks like protons and neutrons. The quantum wave function for two or more identical fermions reverses phase if any two particles are swapped; this leads to the Pauli exclusion principle, which gives a zero probability for two fermions being in exactly the same state. The wave function of a single fermion reverses phase if the particle is rotated by 360 degrees, and is only restored exactly by two full rotations. Fermions have a spin which is an odd-integer multiple of half the fundamental unit of angular momentum. In <a href="#KozuchTheory">Kozuch Theory</a>, all these properties arise from the topology of the particle’s wormhole.</p>

<p class="gloss"><a id="KozuchTheory"></a><b>Kozuch Theory</b>. A provisional unified theory of physics developed in the mid-twenty-first century. Kozuch Theory describes the universe as a ten-dimensional <a href="#fiberbundle">fiber bundle</a>; its size in six dimensions is sub-microscopic, so only the familiar four dimensions of space-time are immediately apparent. Particles such as electrons are actually the mouths of very narrow wormholes, an idea first suggested by the twentieth-century physicist John Wheeler. Renata Kozuch developed a model in which the properties of different particles are due to the different ways wormhole mouths can be connected in the six extra dimensions.</p>

So, WTF?

[Sella174]

On page 21 of the Amazon Kindle Publishing Guidelines 2017.4 it states:

To avoid unintentional footnote pop-ups, internal links that are not footnotes should not be formatted with bi-directional hyperlinks.

[AlanHK]

So, since the boson and fermion paras each refer to the other, they decide I want them to be footnotes.

Another great example of a program assuming you're too stupid to know what you want so they do what they think you want regardless.

Anyone know a way to defeat this, aside from removing the links?
Maybe if the anchor and the "back link" were in different paras, it would block this. Maybe I could put the link before the paragraph, in a div if necessary for syntax. Will try that.

Or, in this book I don't have any intended footnotes, so a way to turn them off entirely would work.

[AlanHK]

Tried separating the links from the text:

<p><a id="boson"></a></p>
<p class="gloss"><b>boson</b>.....

Tried both div and paras. Both looked fine (no space taken by the link para), and worked in epub. But on Kindle, they were still footnotes. even though the links are pointing to the anchor paragraphs and not the following text para.

Amazon no doubt again thought this is saving me from my dumb formatting.

Tried putting t=some text in the link para:

<p><a id="fermion"></a>BLAH</p>
<p class="gloss"><b>fermion</b>. All elementary particles

Still popup footnotes.

[Sella174]

Not sure if this will solve the problem, but in the aforementioned guide under Section 14, dictionaries are discussed.

I do not see what is happening as being Amazon's fault. Rather I would (and am) blaming the W3C for derailing HTML. For how long have people been screaming for a <footnote> tag? Ten, twenty years? Forget it, because social media and games get preference and they have no use for footnotes, endnotes, summaries, etc.

As an aside, for correctly structured HTML, your glossary should be in the form of a definition list (<dl>). I see that Kindle understands it.

[AlanHK]

Originally Posted by Sella174

Not sure if this will solve the problem, but in the aforementioned guide under Section 14, dictionaries are discussed.

It's a novel with a glossary of technical terms, so no.

I do not see what is happening as being Amazon's fault. Rather I would (and am) blaming the W3C for derailing HTML. For how long have people been screaming for a <footnote> tag? Ten, twenty years? Forget it, because social media and games get preference and they have no use for footnotes, endnotes, summaries, etc.

Of course it's their fault: they create a footnote from my text because they imagine that's what I might be doing, not because I asked for one. Their guidelines have "Amazon requires formatting footnotes with bi-directional hyperlinks". Which is fine. But then the arrogant jerks make the reverse also true. I can see their motive to automagically enable older books with bidirectional footnotes to use popup footnotes. But they provide no way to override their assumption.

They say: "Non-footnote links should use the format A links to B and B links to C instead."
As above, I tried that and it did not work.

As an aside, for correctly structured HTML, your glossary should be in the form of a definition list (<dl>). I see that Kindle understands it.

Semantically, yes, and looks fine on a monitor, but on a portrait 6" screen not so great. It puts the term on its own line, and indents the definition, wasting a lot of space on a small page. There would be much less text per screen.
I care about how it looks, not how a hypothetical bot might parse it.

I was using this for style:

.gloss {
    text-indent: -1.5em;
    margin-top: 0.5em;
    margin-left: 1.5em;
}

[Tex2002ans]

Does it act any differently if you attach the ID to the <p>?

Spoiler:

  <p class="gloss" id="boson"><b>boson</b>. All elementary particles can be classified as either bosons or <a href="#fermion">fermions</a>; the bosons include photons and gluons. The quantum wave function for two or more identical bosons is unchanged if any two particles are swapped, and the wave function for a single boson is unchanged if the particle is rotated by 360 degrees. Bosons have a spin which is an integer multiple of the fundamental unit of angular momentum. In <a href="#KozuchTheory">Kozuch Theory</a>, all these properties arise from the topology of the particle’s wormhole.</p>

  <p class="gloss" id="fermion"><b>fermion</b>. All elementary particles can be classified as either <a href="#boson">bosons</a> or fermions; the fermions include electrons and quarks, and composites of three quarks like protons and neutrons. The quantum wave function for two or more identical fermions reverses phase if any two particles are swapped; this leads to the Pauli exclusion principle, which gives a zero probability for two fermions being in exactly the same state. The wave function of a single fermion reverses phase if the particle is rotated by 360 degrees, and is only restored exactly by two full rotations. Fermions have a spin which is an odd-integer multiple of half the fundamental unit of angular momentum. In <a href="#KozuchTheory">Kozuch Theory</a>, all these properties arise from the topology of the particle’s wormhole.</p>

  <p class="gloss" id="KozuchTheory"><b>Kozuch Theory</b>. A provisional unified theory of physics developed in the mid-twenty-first century. Kozuch Theory describes the universe as a ten-dimensional <a href="#fiberbundle">fiber bundle</a>; its size in six dimensions is sub-microscopic, so only the familiar four dimensions of space-time are immediately apparent. Particles such as electrons are actually the mouths of very narrow wormholes, an idea first suggested by the twentieth-century physicist John Wheeler. Renata Kozuch developed a model in which the properties of different particles are due to the different ways wormhole mouths can be connected in the six extra dimensions.</p>

[Sella174]

Originally Posted by Tex2002ans

Does it act any differently if you attach the ID to the <p>?

I think what AlanHK is trying to do is make it possible for the reader to quickly jump to the glossary and then back to the text. I must admit, for that back-links are essential ... unless the Kindle has some dynamic bookmark method that can be set when going to the glossary, kind of like a finger at your place in the paper book.

[Tex2002ans]

Originally Posted by Sella174

I think what AlanHK is trying to do is make it possible for the reader to quickly jump to the glossary and then back to the text.

I would recommend rereading the initial post + looking at the code samples themselves.

The issue is that there are multiple types of links:

• Type A: One-Way Link (these work as intended)
  • boson -> Kozuch Theory
  • fermion -> Kozuch Theory
• Type B: One-Way Links which happen to point to each other later on
  • boson -> fermion
  • fermion -> boson
• Type C: "Bidirectional" Links
  • All the different types of "footnote code" you can come up with

Amazon's heuristics are just thinking the AlanHK's Type B is a very poorly coded Type C. So instead of jumping you to that point in the book, it tries to be "helpful" and treat it as a pop-up footnote instead.

Originally Posted by AlanHK

Of course it's their fault: they create a footnote from my text because they imagine that's what I might be doing, not because I asked for one. Their guidelines have "Amazon requires formatting footnotes with bi-directional hyperlinks". Which is fine. But then the arrogant jerks make the reverse also true. I can see their motive to automagically enable older books with bidirectional footnotes to use popup footnotes. But they provide no way to override their assumption.

This would be a case where a properly coded EPUB3 with epub:type="footnote" would be able to tell the parser what you meant or didn't mean... but the thing is, the vast majority of books are/were coded like garbage and aren't marked up semantically (which is why Amazon created those footnote heuristics in the first place).

What happens if the ID you are jumping to isn't in the same <p> as the one-way links?

Spoiler:

<p id="boson"><b>boson</b></p>

<p>All elementary particles can be classified as either bosons or <a href="#fermion">fermions</a>; the bosons include photons and gluons. The quantum wave function for two or more identical bosons is unchanged if any two particles are swapped, and the wave function for a single boson is unchanged if the particle is rotated by 360 degrees. Bosons have a spin which is an integer multiple of the fundamental unit of angular momentum. In <a href="#KozuchTheory">Kozuch Theory</a>, all these properties arise from the topology of the particle’s wormhole.</p>

<p id="fermion"><b>fermion</b></p>

<p>All elementary particles can be classified as either <a href="#boson">bosons</a> or fermions; the fermions include electrons and quarks, and composites of three quarks like protons and neutrons. The quantum wave function for two or more identical fermions reverses phase if any two particles are swapped; this leads to the Pauli exclusion principle, which gives a zero probability for two fermions being in exactly the same state. The wave function of a single fermion reverses phase if the particle is rotated by 360 degrees, and is only restored exactly by two full rotations. Fermions have a spin which is an odd-integer multiple of half the fundamental unit of angular momentum. In <a href="#KozuchTheory">Kozuch Theory</a>, all these properties arise from the topology of the particle’s wormhole.</p>

If you want the term on the same line, maybe you can try do some CSS float magic.

Or maybe the <dl> might tell the heuristics "this ain't no damn footnote".

In this case, you know that your glossary code is correct, and it is just the heuristics that are guessing poorly.

I would be interested to test the Amazon footnote heuristics to its limits, but I don't own a Kindle. There has to be some distance and criteria where the algorithm says "nope, that's not a footnote".

Although I suspect a non-standard solution to try to get around the heuristics would be very hackish, and could break with a future firmware update.

Originally Posted by Sella174

unless the Kindle has some dynamic bookmark method that can be set when going to the glossary, kind of like a finger at your place in the paper book.

This is a separate issue.

On your typical ereader, you may only get one/none "back". It is very easy to move forward, and go from:

fermion -> boson -> Kozuch Theory -> fiber bundle

But if you pressed back, you may go from fiber bundle -> Kozuch Theory. If you press back again, you would jump back to the device's Library/Home Screen.

On something like a browser, they typically allow you to stack tens/hundreds of "backs". So you would easily be able to jump back/forward from:

fermion <-> boson <-> Kozuch Theory <-> fiber bundle <-> [...]

Side Note: Hitch has also written extensively about a semi-related problem, "Many to One" links. It mostly comes up with Indexes, but it seems like it could be a serious issue in Glossaries as well.

[Sella174]

I had a think about this: the solution is to use the <dl>, <dt> & <dd> markup.

[AlanHK]

Originally Posted by Tex2002ans

What happens if the ID you are jumping to isn't in the same <p> as the one-way links?

<p id="boson"><b>boson</b></p>

<p>All elementary particles can be classified as either bosons or <a href="#fermion">fermions</a>

Tried that; the fermion link still becomes a footnote.
Same with divs. Presumably if I put enough stuff between the link and the body eventually it would fail the heuristic.

On your typical ereader, you may only get one/none "back". It is very easy to move forward, and go from:
fermion -> boson -> Kozuch Theory -> fiber bundle
But if you pressed back, you may go from fiber bundle -> Kozuch Theory. If you press back again, you would jump back to the device's Library/Home Screen.

On Kindle you get many backs, haven't yet run into a limit.

This particular book uses some jargon, some real, some invented, and had a glossary at the back.
Kindle lets you select any text and search within the book, gives you a list of hit links and few lines of text. I found myself having to page to the end of the hits (some words used many times so several pages of hits) to remind me of the defn.

So I moved the glossary to the front so the glossary links were always the first search hits. Then I noticed there were "See" references in the glossary and coded them as links. Then went whole hog and coded less explicit cross refs; thus running into the issue at hand.

Works fairly well, even with the footnote glitch. Can live with that if the cost of breaking this auto-footnote thing is excessive complication of the code; at least for personal use.

[najgori]

in the last publishing guidelines amazon defines "preffered" method for footnotes as:
...
<aside id="ft-1-1" epub:type="footnote">
...
have you tried to explicitly declare link as something other than footnote (from the list)?
...maybe it'll work

[jcsalomon]

Originally Posted by AlanHK

Semantically, yes, and looks fine on a monitor, but on a portrait 6" screen not so great. It puts the term on its own line, and indents the definition, wasting a lot of space on a small page. There would be much less text per screen.
I care about how it looks, not how a hypothetical bot might parse it.

That’s where CSS comes in:

dt {
    float: left;
    clear: left;
    font-weight: bold;
}
dt:after {
    content: ". ";
}

<dl>
<dt id="boson">boson</dt><dd>All elementary particles can be classified as either bosons or <a href="#fermion">fermions</a>; etc., etc., etc.</dd>
<dt id="fermion">fermion</dt><dd>All elementary particles can be classified as either <a href="#boson">bosons</a> or fermions; etc., etc., etc.</dd>
</dl>

[Hitch]

Originally Posted by Tex2002ans

Side Note: Hitch has also written extensively about a semi-related problem, "Many to One" links. It mostly comes up with Indexes, but it seems like it could be a serious issue in Glossaries as well.

Well, firstly, IME, everybody wants the parsers/renderers to do what they want, and that's where we always end up in this sort of situation--where AlanHK doesn't want his x-reffed links to be pop-up footnotes, but they ARE. My question is this: is that really a problem? Why not code the back links, and let them ALL pop-up? On those devices that do that, that is. After all, to this day, most don't. Presumably, the book isn't ONLY for his use, right? So...unless there are also footnotes, in this novel, and that would then be confusing, I'd say, let it go and make them ALL popup. FWIW.

Secondly--yes, it's a problem in glossaries, as well. Same exact thing, many to one is ALWAYS a problem when dealing with how to get back. We've had someone persist in INSISTING that multiple links, to a resource that's cited in an endnote, get "back" so now we have links to/from that look like this:

Link 1 from wherever in the book (outbound)
link 2 from wherever in the book (later)(outbound)
link 3 from wherever in the book (yet later)(outbound)

Endnote Resource [1][2][3].

Honestly, I don't think that the typical user will think "oh, right, that's my second/third use of that reference time" and know which of the bracketed links to click.

Spoiler:

Of course, she also is insisting on something else, and I'm 99% sure that users won't know to click those, either (section heads, like "About Butter" or the like. Her body text section heads aree STYLED blue--yes, kids, just like links--so..who the hell's going to think "oh, yeah, that section head is a link, unlike the other 50 that aren't"?

Whatev.

Unless/until there is bonafide separate footnote coding, I don't see how this is dealt with. There are many perfectly valid reasons for backlinks, unrelated to footnotes.

Hitch