MobileRead Forums - View Single Post

poxalew · 11-16-2015, 03:11 AM

Correct, it didn't. Amazon modified it (or, at least, merged stuff from elsewhere) and has open sourced it; I found some spare time and looked into it.

We can compare upstream https://trac.webkit.org/browser/rele.../Hyphenation.h (which is mostly a stub) to Source/WebCore/platform/text/Hyphenation.h from Amazon's tarball. The latter contains a reasonable hyphenation engine. It uses hunspell's dictionaries (telltale "hyph_en_US.dic" and "hyph_en_GB.dic" references in Hyphenation.cpp.) The code is reasonably well documented and quite readable. It's guarded by #ifdef LAB126 / #endif //LAB126 stanzas, so grepping for LAB126 through Amazon's source should yield interesting data.

So it looks like #1 is the case. #2 is right out (the source code dump doesn't even contain a JS interpreter and I doubt Amazon would have rolled their own for the same economic reasons for which they didn't reimplement rendering), and #3 indeed looks not to be the case.

Having the source is interesting in a number of ways, perhaps even for calibre if at some point in the future the JS hyphenation becomes a bottleneck and needs to be replaced with a lower runtime overhead implementation. Amazon's hyphenation implementation is BSD-licensed according to the copyright headers in the source files.

(By the way, my blanket statement about WebKit being LGPL was wrong; it's mostly BSD. LGPL governs just small parts of JavaScriptCore and a part of WebCore, but fortunately that was enough to force the entire WebCore into the open.)

11-16-2015, 03:11 AM	#237
poxalew Junior Member Posts: 9 Karma: 166666 Join Date: Aug 2014 Device: Kindle PW2	Correct, it didn't. Amazon modified it (or, at least, merged stuff from elsewhere) and has open sourced it; I found some spare time and looked into it. We can compare upstream https://trac.webkit.org/browser/rele.../Hyphenation.h (which is mostly a stub) to Source/WebCore/platform/text/Hyphenation.h from Amazon's tarball. The latter contains a reasonable hyphenation engine. It uses hunspell's dictionaries (telltale "hyph_en_US.dic" and "hyph_en_GB.dic" references in Hyphenation.cpp.) The code is reasonably well documented and quite readable. It's guarded by #ifdef LAB126 / #endif //LAB126 stanzas, so grepping for LAB126 through Amazon's source should yield interesting data. So it looks like #1 is the case. #2 is right out (the source code dump doesn't even contain a JS interpreter and I doubt Amazon would have rolled their own for the same economic reasons for which they didn't reimplement rendering), and #3 indeed looks not to be the case. Having the source is interesting in a number of ways, perhaps even for calibre if at some point in the future the JS hyphenation becomes a bottleneck and needs to be replaced with a lower runtime overhead implementation. Amazon's hyphenation implementation is BSD-licensed according to the copyright headers in the source files. (By the way, my blanket statement about WebKit being LGPL was wrong; it's mostly BSD. LGPL governs just small parts of JavaScriptCore and a part of WebCore, but fortunately that was enough to force the entire WebCore into the open.)