Best Practice: Embed Fonts??

[Man Eating Duck]

Originally Posted by Adjust

Also you cannot rule out the Font License, when dealing and embedding font into ePub, because you can't subset, you are effectively distributing full and complete font without having permission to do so. Unless you use free fonts, which the majority of publishers do not...

We do

That's why we use the DejaVu fonts. While not very pretty, they are easy to read on a variety of screens and have very decent Unicode coverage, while being free for commercial use. There are not many fonts with these characteristics, Libertine is an alternative although it has marginally fewer glyphs.

[Adjust]

Originally Posted by Man Eating Duck

We do

That's why we use the DejaVu fonts. While not very pretty, they are easy to read on a variety of screens and have very decent Unicode coverage, while being free for commercial use.

Yeah I think that's the main problem... Most authors who don't have contact with publishing houses or other industry bodies probably won't be aware...

So I think there could be, if the makers of the fonts, which people are embedding decide to go looking, some very unhappy people out there who have embedded fonts without knowing the licensing conditions.

4. Font Embedding.
The basic P22 license allows for the embedding of P22 fonts only if the document created is set to "Print and Preview". If P22 fonts can be extracted, edited and therefore transferred in any way, an additional license is required to account for each recipient of the document and font file(s)

And from Linotype website:

1.5 Embedding of the Font Software into electronic documents or Internet pages is only permitted under the absolute assurance that the recipient cannot use the Font Software to edit or create a new document (read-only). It must be ensured that the Font Software cannot be fully or partially extracted from said documents.

[ozaru]

Originally Posted by shall1028

There is a place for saving space. [...] A quick perusal of my font folder shows 161 files only 60 of which exceed 400K.

I have a problem in that the ePub I've just painstakingly produced (and which displays fine on my PC in Sigil, Calibre and EPUBReader/Firefox), fails totally in Adobe Digital Editions -- because it contains a few Unicode characters (not even on the roadmap for ADE, it appears: see http://forums.adobe.com/thread/314859)

Some might be relatively easy to fix with a smallish font (vowels with macrons, used throughout the text), but there are also about a dozen Japanese characters used on just 2 pages, for which the only solution I can think of is embedding a whole Japanese font file -- around 5MB. Not ideal, huh?

If there were some way to subset this in ePub that would be great, but I've not seen it anywhere. I suppose an alternative would be to 'extract' the characters from a Japanese font whose licensing terms allow it (I bet MS Mincho doesn't) and create a new font from them, just for this one book, which seems a bit silly -- and I suspect will also break various reading systems.

Any advice?

[Man Eating Duck]

Originally Posted by ozaru

If there were some way to subset this in ePub that would be great, but I've not seen it anywhere.

This online service seems to be able to do what you need: http://www.fontsquirrel.com/fontface/generator

Select expert options, where you can enter single glyphs for subsetting with preview. I tried subsetting five glyphs from a ttf font, and ended up with a ttf file only 3.5 KB in size.

[charleski]

Originally Posted by ozaru

Some might be relatively easy to fix with a smallish font (vowels with macrons, used throughout the text), but there are also about a dozen Japanese characters used on just 2 pages, for which the only solution I can think of is embedding a whole Japanese font file -- around 5MB. Not ideal, huh?

Subsetting, as mentioned above, is a good route, though you still have to deal with licensing issues. If it's only a dozen characters on 2 pages, why not just render them to png files and use those? All font licences will allow that. You can get the images to show up in-line with the text.

Macrons are a different issue, as most fonts have a macron, but rely on overstriking to use it. You can get an epub to overstrike by using a span with adjusted margins, eg
o<span class="o-macron">& #175;</span> {remove the space}
with
span.o-macron {
margin-left: -0.37em;
margin-right: 0.12em
}

But frankly this is very fiddly, needs to be fine-tuned for each character and breaks if the font is changed. For macrons it would be easier to embed something like SIL Gentium Basic which includes macroned vowels and has a fully open licence that allows epub embedding. It's still a hefty font at over 200k for each weight and style, but you could get away with not having to subset. (A lot of people use SIL's Charis, but personally I can't stand slab-serifs, and its massively inclusive range means each weight is 1.6MB.)

[ozaru]

Originally Posted by charleski

Subsetting, as mentioned above, is a good route, though you still have to deal with licensing issues.

True, but even free & amateur fonts that 'have problems' in certain areas may be acceptable for those few characters, so I think there is likely to be a good choice of licensable material.

Originally Posted by charleski

If it's only a dozen characters on 2 pages, why not just render them to png files and use those?

I did consider that, but don't really like that approach on principle -- it's like translating a web page into GIFs (and losing the ability to search, reflow nicely at line ends, output to voice readers, etc.).

Originally Posted by charleski

Macrons are a different issue, as most fonts have a macron, but rely on overstriking to use it.

Those types never looks quite as good as e.g. U+014D -- that method also increases the size, and makes searching awkward.

Originally Posted by charleski

For macrons it would be easier to embed something like SIL Gentium Basic which includes macroned vowels and has a fully open licence that allows epub embedding. It's still a hefty font at over 200k for each weight and style, but you could get away with not having to subset.

Alternatively, I could use it and subset to strip out all the characters I don't need, decreasing the size... sounds tempting. I suppose I could also get around it by specifying e.g. Times New Roman on the basis that most systems will have it available, including macrons? (Currently I simply put 'serif', and maybe that's why Adobe Digital Editions fails. Mind you, it also fails to pick up on several things in the CSS so I'm not sure how reliable that will be.)

[charleski]

Originally Posted by ozaru

I suppose I could also get around it by specifying e.g. Times New Roman on the basis that most systems will have it available, including macrons?

They won't. Mobile ADE-based readers only need to include a font with glyphs that cover the subset specified by Adobe.

[ozaru]

Originally Posted by Man Eating Duck

This online service seems to be able to do what you need: http://www.fontsquirrel.com/fontface/generator

It did indeed look promising -- I managed to get my 25 characters from a free Japanese 6MB font down into a tiny 9K TTF. In Charmap the character spacing was a bit off, but in Sigil it appeared to work OK (although Sigil was displaying the characters OK before in any case, so I'm not 100% certain...). In Adobe Digital Editions, however, they rigidly stayed as "?" marks. I note some similar comments in the ThreePress blog. But even on a Japanese website they demonstrate ADE failing to display Japanese ePubs. I wonder if changing the TTF to OTF would make any difference... Yet another avenue to explore.

Microsoft's Expression Studio (trial) also apparently has the facility to subset a font. I'm not too confident it will work if Squirrel failed, but the options are running out... Perhaps I ought to give up on ePub for now and try Kindle instead.

[ozaru]

Originally Posted by ozaru

In Adobe Digital Editions, however, they rigidly stayed as "?" marks [...] even on a Japanese website they demonstrate ADE failing to display Japanese ePubs

OK, scrub that: I've now found another Japan-based website that demonstrates the only change needed is adding xml:lang="ja" after xmlns="http://www.w3.org/1999/xhtml" -- then the Japanese characters (and even macrons) do display in ADE.

Embedding therefore seems irrelevant (although I have to say I've not tested it in ADE on a system with no Unicode fonts). It's not a perfect solution yet, though, as the font for that whole chapter has now changed to an awful Japanese typeface, completely ignoring what is specified in the CSS: indeed, ADE appears to ignore my CSS entirely. That probably calls for a separate thread though. Thanks to all who assisted.

[JSWolf]

The font most publishers use is Charis. It's not a bad font overall. It definitely is better then the default serif font used in ADE. And you will get true bold and true italics. So that's good for sure.

[Adjust]

Originally Posted by JSWolf

The font most publishers use is Charis.

That's the font Indian companies (CodeMantra etc) use. It isn't a conscious decision on the publishers end...

[marytg]

Embedded fonts will be allowed in EPUB3 which is due out soon. This is quote from idpf.org site:

3.4 Fonts
EPUB 3 does not require that Reading Systems come with any particular set of built-in system fonts. As occurs in Web contexts, Users in a particular locale may have installed fonts that omit characters required for other locales, and Reading Systems may utilize intrinsic fonts or font engines that do not utilize operating system installed fonts. As a result, the text content of a Publication may not natively render as intended on all Reading Systems.

To address this problem, EPUB 3 supports the embedding of fonts to facilitate the rendering of text content, and this practice is recommended in order to ensure content is rendered as intended.

Support for embedded fonts also ensures that Publication-specific characters and glyphs can be embedded for proper display.

This is what an Adobe staff person said about embedding UNICODE glyphs in ADE. It seems they feel the same about Cyrillic fonts:

Full Unicode font support for all glyphs would bloat the Digital Editions download from ~6Mb to around 32 to 40 Mb. This is not something we are anxious to do. We are looking at other mechanisms to download the fonts on demand, but this is fairly low in priority compared to long laundry list of other features we want to do.

Understandable, but reflective of American arrogance in my opinion.

[JSWolf]

Embedded fonts is allowed in ePub now. No need to wait until ePub 3 is is finished.

[Jellby]

Originally Posted by marytg

This is what an Adobe staff person said about embedding UNICODE glyphs in ADE. It seems they feel the same about Cyrillic fonts:

Nonsense. All they need is allow using system fonts. Web browsers don't come with a full unicode font embedded either, and they don't have that problem.

[marytg]

ADE doesn't display Cyrillic characters even if the UNICODE code is in the epub. For example, when it encounters the code for the letter "č" it displays a "?." A sad state of affairs for lots of Europeans.

Embedding fonts in an epub seems to increase the file size greatly but if that's what we have to do to properly display non-English characters, so be it.