Quote:
Originally Posted by jhowell
I disagree with whoever wrote that in the wiki. That was early speculation, since disproved.
The original post also mentioned koreader. As far as I know that doesn't support KFX. I don't know whether or not it supports hyphenation some other way.
|
Yes, I can appreciate that Amazon would want to avoid inserting (many) characters into the content stream. So they compute offsets into that stream (representing positions where hyphens can be) and store those in a separate stream, and they can merge these very efficiently as the content stream is read into memory and page layout is applied.
The computation is done in the cloud (once for each title, and re computing whenever they update the hyphenation behavior), obviating the need for (at least several) hyphenation dictionaries on the device, eliminating CPU usage to do hyphenation lookups, and also lets them refine the hyphenation behavior without any device updates.