Thread: AZW Conversion Output Plugin
View Single Post
Old 07-29-2011, 09:07 PM   #19
nickredding
onlinenewsreader.net
nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'nickredding knows the difference between 'who' and 'whom'
 
Posts: 328
Karma: 10143
Join Date: Dec 2009
Location: Phoenix, AZ & Victoria, BC
Device: Kindle 3, Kindle Fire, IPad3, iPhone4, Playbook, HTC Inspire
Quote:
Originally Posted by kovidgoyal View Post
Well it isn't the TBS bytes (at least not just the TBS bytes). I've decoded them almost entirely and having the same set as produced by kindlegen 1.2 seems to make no difference. I suspect its going to need some special markup as well.
Are they EXACTLY the same? I found that Kindlegen inserts data bytes into SOME of the TBS records that don't seem to make sense (and I can't figure out the logic that generates them), but if they are omitted--bingo, the K3 problem.

Interestingly, if you look at Amazon-generated periodicals (e.g. the New York Times) they have predictable TBS records (without the unpredictable TBS bytes) but those files have an additional data field in SOME of their NCX entries that doesn't seem to make sense (like the extra TBS bytes for Kindlegen files). If you generate files with identical TBS records to the Amazon-generated periodicals, but without the nonsensical NCX data bytes--bingo, the K3 problem.

So I believe there are two issues: Kindelgen does something weird in the TBS records and Amazon's periodical publishing system does something weird in the NCX entries. If you generate a file that has NEITHER of these weird characteristics--bingo, the K3 problem.
Quote:
And I've discovered kindlegen 1.2 can produce periodicals with all the index records and TBS data that do not work on the Kindle 3. Why am I not surprised.
OK, that's really interesting. I haven't seen that, but it should be noted that Amazon took some stuff OUT of Kindlegen between 1.1 and 1.2 that looks like it was intended to frustrate people using it to generate periodicals (e.g., the masthead processing disappeared). Instapaper is still using 1.1, and it would be interesting to see if 1.1 produces the same defective result as 1.2 on the source you are using.

I too thought it would be a relatively straightforward reverse-engineering job to fix the K3 problem. However, after 150 hours of work I concluded that the folks at Amazon are either really bad software developers or they are trying to make life difficult for people using alternate methods of generating periodicals (that don't involve paying Amazon for subscriptions). My money is on the latter.
nickredding is offline   Reply With Quote