23 October 2014

Ebooks for Reform Jews

Since I've been trying, mostly without success, to get work creating high-quality ebooks of interest to Reform Jews, I was curious to survey the state of that business.

I decided, somewhat arbitrarily, that I would use the following data as a proxy for the state of ebooks of interest to Reform Jews.

I took as my sample the list of previously recommended non-fiction books from the URJs list of Significant Jewish Books.

Then, for each of these books, I created a row in a Google Spreadsheet with the following data.
  1. The Amazon sales rank for the best-ranking paper version of that book.
  2. The Amazon sales rank for the Kindle version of that book, if it existed.
To do the above, I used a combination of Amazon's pages and Aaron Shepard’s Sales Rank Express.

Of course, this data is only a proxy for what I really want to get at. I am using URJ's list as a proxy for the vague (and much broader) notion of "books of interest to Reform Jews." And, less dubiously, I am using Amazon paper and Kindle sales as a proxy for all paper and ebook sales.

This data collection turned out to be surprisingly tricky in many cases. Amazon isn't called Amazon for nothing: it is a jungle of not-that-well-organized data. In particular, for many books, it is not clear whether a Kindle edition exists, because a link to the Kindle edition is not provided from some or all of the pages showing a paper edition. In one case, the Kindle edition was only available for a previous edition of the book, which had a slightly different title! So you have to do a careful, separate search for a Kindle edition if one seems to not exist.

Anyway, enough of my kvetching, and on to the results.

Kindle editions are available for 31 out of the 50 books (62% of the books). As to what the quality of these Kindle editions is, I have my doubts. In my experience, Kindle ebooks vary greatly in quality, but all are lower quality (e.g. have more typos) than their paper counterparts.

Somewhat surprisingly, there doesn't seem to be a strong relationship between the rank of the paper book and the presence of an ebook. I would have expected publishers to be more pragmatic in their choice of which titles they have chosen to convert from their back catalog, letting worse-ranked titles languish in paper-only obscurity.

The best-ranking paper book without an ebook was Primo Levi's The Periodic Table. Its paper rank was 30,845.

On the other hand, of the 15 books whose rank was worse than (i.e. higher than) 500,000, about half (8 of them) were available as ebooks!

Perhaps low-quality ebook conversion is so cheap that publishers don't worry about a paper book's low performance too much in deciding whether to convert it to be an ebook. What factors they do consider in making the decision, I would love to know.

In four cases I deem significant, ebooks actually out-rank their paper counterparts. This comparison is a little arbitrary since I don't think ebook and paper book rankings can be directly compared, since there are many more paper books for sale there than there are ebooks for sale. Even if there were exactly the same number of ebooks and paper books for sale, it would only tell you whether the ebook sells better, relative to all ebooks, than the paper book sells, relative to all paper books. In other words it would not tell you whether the ebook sells more copies than the paper book.

Still, for lack of a better metric, comparing ebook rank to paper book rank is interesting.

In my spreadsheet, I've used orange as the background color for all cases in which the ebook has a better rank than its paper counterpart.

I only consider the four best-ranked cases of this to be significant since I assume that down in the poor-performing end of the rank distribution (let's say, rank worse than (i.e. higher than) 500,000), there is a lot of noise.

17 October 2014

More on bitonal (inkscale) images

Here are a four updates to my previous post.

  1. Rename: "inkscale" instead of "bitonal"
  2. Web icons show the demand for inkscale
  3. Harry Potter examples
  4. Proposals to standards committees
1. Rename: "inkscale" instead of "bitonal." The mental model is a single-color printing process, i.e. one color of ink on some color of paper. I also considered, but rejected, "foreground-scale" and "currentColor-scale."

2. Web icons show the demand for inkscale. In my previous post, I mentioned two ways of implementing inkscale using today's standards. One way is fonts, and the other way is SVG. What I did not mention is that this is not just a theoretical capability: people are widely implementing icons for their websites using both of these technologies.

Nonetheless, I can't help but feel that the font approach is fundamentally a hack, not a real, long-term solution. And, while the SVG approach feels better, it fails to address raster images. This failure is not much of a problem in the realm of icons, since raster icons are probably best avoided anyway!

3. Harry Potter examples. There are many cases where you would not want to use inkscale. For example, most grayscale photographs should be rendered as grayscale, not inkscale. But there are many drawings for which the decision requires a careful judgment call.

Take, for example, the following screen snip from the beginning of a Harry Potter chapter:
Though this doesn't look great, the problem is only made worse by converting to inkscale (which, in this case, means inverting):
To reaffirm your faith that there are many examples where inkscale is a clear win, here's another screen snip:
And here it is with its two images converted to inkscale:
Here's my guess at a general guideline for whether to convert to inkscale or not: If the intensity values of the image have meaning, don't pervert this meaning by converting to inkscale.

For example, the Hagrid image above is not a line drawing: it uses intensity (shading) to indicate something about the implied color of its objects and the implied source(s) of lighting. It implies that his boots are dark, and that they are lit from the left.

Whereas, in the letter, the swoosh and the signature are not real, lit objects. Or rather, we are to imagine them as real objects, but we do not seek to represent them as such.

One rough version of the guideline would be the following. If the image should really have been represented in a vector format like SVG, it should probably be inkscale. Whereas, if it is appropriately represented in a raster format, it should probably stay grayscale. I'm sure there are many exceptions to this, e.g. an elaborately shaded SVG image.

Even though it takes us beyond the ideas of inkscale, I can't resist suggesting that it could be an interesting added value to the ebook if the signature appeared in a slightly different color than the text. Doing this in the paper book would presumably introduce a big increase in marginal cost, but of course in an ebook, it would have only the fixed cost of figuring out how to implement it. To satisfy the color-scheme flexibility that is in the spirit of inkscale, we might like to define the color of the signature as a slightly hue-rotated version of the text color. If the text color has no hue, i.e. is black or white, we might give it a bit blue hue. This way, in the standard black (ink)-on-white (paper) color scheme, the signature would appear to be in dark blue ink.

4. Proposals to standards committees. I've submitted the inkscale idea for consideration as part of the next revision of EPUB and for consideration for inclusion in the CSS filter effects module.

02 October 2014

What would Isaiah do?

[This was originally a post I made on facebook in response to the article, "Orthodox Man Refuses To Sit Next to Feminist Activist on Airplane."]

Christians sometimes ask themselves, "What would Jesus do?" In what could be thought of as a Jewish analogy to this question, Heschel challenges all Jews to ask themselves: would your Judaism be intelligible to Isaiah [1]? In other words, "What would the Hebrew prophets have you do?"

The Hebrew prophets teach us that, in analogy to God's limitless concern for us, there should be no limit to our concern for others.

This means that mitzvot should never be observed in isolation of concern for others.

Thus, even if I grant that avoiding contact with a member of the opposite sex is a mitzvah, it cannot be observed in isolation, i.e. it cannot be observed without considering how it causes others to feel.

We are approaching Yom Kippur. Consider what Isaiah has to say, in the haftarah for YK morning, about the mitzvah of fasting, if observed in isolation of concern for others:
58:5. Is this the fast I have chosen?
A day of self-affliction?
Bowing your head like a reed,
and covering yourself with sackcloth and ashes?
Is this what you call a fast,
a day acceptable to the Eternal? 
6. Is not this the fast that I have chosen:
to unlock the shackles of injustice,
to loosen the ropes of the yoke,
to let the oppressed go free,
and to tear every yoke apart? 
7. Surely it is to share your bread with the hungry,
and to bring the homeless poor into your house;
when you see the naked, to cover them,
never withdrawing yourself from your own kin. 
8. Then shall your light break forth like the dawn,
and your healing shall quickly blossom;
your Righteous One will walk before you,
the glory of the Eternal will be your rear guard. 
9. Then, when you call,
the Eternal will answer;
when you cry, God will say: Here I am.
(Translation by Chaim Stern.)

Particularly resonant with the airplane issue is Isaiah's idea of "never withdrawing yourself from your own kin."

What would Isaiah have done, had he been on that plane?


1. The quote from Heschel I have in mind is the following, from "To Be a Jew: What Is It?" in Moral Grandeur and Spiritual Audacity, p. 9.
Our way of life must remain to some degree intelligible to Isaiah and Rabbi Yochanan ben Zakkai, to Maimonides and the Baal Shem.

Condemnation and Consolation

[This is something I wrote last year that appeared in my temple's newsletter. It was in response to a request for comments on a Yom Kippur sermon that many temple members found provocative: some in a good way, some in a bad way.]

Though we won't have another prophet until The Anointed One comes, we must keep the prophetic voice alive.

This voice, as typified by Isaiah, is a bewildering combination of condemnation and consolation.

It may seem to contain contradictions, in making us responsible for so much of what is wrong with the world, while reminding us that there are limits to what we can control.

Perhaps the most important reminder of our limits is Shabbat.

Consider the Haftarah for Yom Kippur Morning, from Isaiah.

As is the prerogative of prophets, Isaiah speaks for God when he asks, in apparent disgust, "Is this the fast I have chosen?"

The answer is of course "no." The fast he has chosen is not just refraining from eating on one day. The fast he has chosen is also refraining, every day, from the self-indulgence that prevents us from caring for others. The hardest fast of all. The one that seems to place all the responsibility on us.

But, the portion concludes by reminding us that to be rewarded we must not only fix the world, but also "keep from trampling the Sabbath."

So we must not only fix the world, but also be mindful that we cannot fix it all. There is a remainder (one seventh?) that only God can and will fix.

We are offered the additional consolation that when all of this is done, we will ride high and our fast will be over:
I will cause you to ride upon the heights of the earth,
and I will feed you with the portion of Jacob your father
–The Eternal One has spoken.
This year (5774), Yom Kippur (Shabbat Shabbaton) fell on Shabbat.

Yom Kippur seems particularly well-suited to the condemning part of the prophetic voice, while Shabbat is more suited to the consoling.

What's a rabbi to do, in composing a sermon for such a day?

15 September 2014

HTML needs bitonal images

The only purpose of this post is to preserve links to the following post, whose title I changed, without realizing that a title change causes a URL change!

The newly-title post is here: http://dencklatronic.blogspot.com/2014/09/the-web-needs-bitonal-images.html

11 September 2014

The web needs bitonal images

Web browsers are missing an important feature: bitonal images.

For my purposes at least, I define a bitonal image as a uni-dimensional image, whose scale varies from an undefined background color to an undefined foreground color. So, it is like a grayscale image, but instead of varying from black to white, it varies from an undefined background color to an undefined foreground color.

I'm also assuming that a bitonal image feature would support transparency, in which case it would support a scale that varies from transparent to an undefined foreground color.

The basic problem is that without bitonal images, the colors used in a page's uni-dimensional images cannot be kept in sync with the page's surrounding text colors.

Here, by "uni-dimensional images," I mean images that vary from a fixed background color (or from transparent) to a fixed foreground color, even though these images may be represented multidimensionally, as in RGB. In other words, by "uni-dimensional images," I mean images that could be represented as bitonal, but they are not, since no such representation exists (to my knowledge).

This brings me to an important point. Bitonal images could be implemented by extending an existing image format, in which case no extension to HTML would be needed, though web browsers would need to be upgraded to interpret this new feature of, for example, PNG.

Alternately, bitonal images could be implemented by extending HTML, perhaps as a new attribute of the IMG element that would instruct the browser to "deliberately misinterpret" a grayscale image as a bitonal image. (By "HTML" I mean to include CSS. I.e., I am agnostic as to whether the bitonal image feature, if it does not reside in an image format, should reside in HTML or CSS.)

My particular motivation for wanting bitonal images is ebooks. Ebooks readers are basically specialized web browsers, since the two most commercially important ebook formats are the following.
  • EPUB, which is based on HTML
  • Kindle (MOBI/KF8), which is usually generated from EPUB
The reason why bitonal images are particularly important for ebooks is that, unlike most web browsers and web sites, many ebook readers allow the user to set the foreground and background colors.

So whereas the lack of bitonal images on web pages is primarily an inconvenience for web designers, the lack of bitonal images in ebooks is an inconvenience for end users. For example, consider these following two different "statements of pain."
  • Web designer: "Argh! We can't redo the color scheme without re-generating our images!"
  • Ebook end user: "Argh! I like black-on-beige text but it makes this book's images stick out like a sore thumb, since they are stuck as black-on-white!"
Another reason that the lack of bitonal images is particularly important for ebooks is that many ebooks encode non-Latin characters as images, since unfortunately for a long time many ebook readers did not have Unicode support. It is particularly jarring when such "text-only images" do not match the foreground and background colors of the surrounding text.

Here are some examples of problems created by the lack of bitonal images in ebooks. These are screen snips from the Kindle for PC ebook reader.  I show each example two ways. First, I show it with a problematic color scheme (white-on-black or brown-on-beige) and then I show it with the black-on-white color scheme, in which case there is no problem.

Finally, a few random concluding notes and questions.

A "shout-out" is due to AutoCAD, since it seems to implement bitonal images.

SVG inside HTML already supports bitonal images, for vector elements, i.e. strokes and fills. E.g. an SVG element's "fill" property can be set to the SVG variable "currentColor," which is the current HTML/CSS foreground (text) color. What is not clear to me is whether SVG can be coaxed to transform raster images according to currentColor.

Fonts can, somewhat perversely, be thought of as special-purpose bitonal image formats. In some sense, this is what has created the whole problem: if fonts weren't bitonal, we wouldn't need bitonal images to match text colors!

But, on the flip side, fonts can, somewhat perversely, be used to allow bitonal images in HTML today. The idea is, somewhat perversely, to encode the images that need to be bitonal as glyphs in a custom font. (This may not be so perverse if the images are in fact being used to overcome character set limitations, e.g. spotty Unicode support.)

The End.

Update: a friend who shall remain nameless found the following Stack Overflow discussion of how to implement something close to a bitonal images feature:

The closest thing to bitonal images, at the moment, seems to be the CSS "filter" property, e.g. as described here:


The "filter" property (in its webkit-specific form) is intriguingly demonstrated here:

and here:

I may be missing something, but it seems like the "filter" property, though offering something close to bitonal image support, isn't there yet.

In particular, the "filter" property offers grayscale() and sepia(), but bitonal image support (at least as I have defined it) would require something like foreground(). I.e. grayscale() and sepia() are mapping the image along two specific background/foreground scales, rather than mapping the image along a "current background" to "current foreground" scale.

30 June 2014

Ebooks from Sources

Here’s a link to a minimal, work-in progress web site advertising my ebook consulting services: http://bdenckla.github.io/efs/.