How are epub page numbers set?

[Ripplinger]

Some of those spans call up italics or bold, so to do it right, each one really should be checked and not blindly run a regex script removing them all. Quite a few of them are useless though, even with many <span></span> for no reason multiple times with a sentence.

And with over 10,000 instances of span in the book, I'm not sure I want to do the publisher's job on it since I've read it already. If it was a book I scanned, even for my own use, I'd have gladly cleaned it all up. I did make a lot of corrections for spelling and grammar and works omitted, run together, punctuation mid-sentence where you know it was picked up by OCR and then not proofed very well.

I may still give in though and do it when I'm bored, just to see if the page numbers change still (although I'm not sure they will).

[JSWolf]

Well, Modify ePub will get rid of those empty <span></span> combinations. That's a start. Then you can see what else needs to go.

[Ripplinger]

I made a dent in it last night and got the number down half to a just over 5000 now in a few minutes using regex to remove the <span></span> and <span>&nbsp;</span> there for no reason. I might tackle the rest later this weekend.

That still didn't change the size of the book in size though, not even by 1kb which I find odd. And I still don't see anything that will give the inflated page numbering.

[j.p.s]

Originally Posted by Ripplinger

I made a dent in it last night and got the number down half to a just over 5000 now in a few minutes using regex to remove the <span></span> and <span>&nbsp;</span> there for no reason. I might tackle the rest later this weekend.

That still didn't change the size of the book in size though, not even by 1kb which I find odd. And I still don't see anything that will give the inflated page numbering.

That is why the (false) claim is made that bloat in XML doesn't matter.

The <span></span> probably frequently followed and/or preceded another frequently occurring string. Make two files with 100 and 1000 lines, all of which are:
<prebloat></prebloat><postbloat></postbloat>

Make another couple of files with the same number of lines, all of which are:
<prebloat></prebloat><span></span><postbloat></postbloat>

zip all four files into separate zip files.

ls -l span-*
-rw-r--r-- 1 me me 4500 Jul 26 14:38 span-no.txt
-rw-r--r-- 1 me me  231 Jul 26 14:39 span-no.zip
-rw-r--r-- 1 me me 5800 Jul 26 14:35 span-yes.txt
-rw-r--r-- 1 me me  245 Jul 26 14:39 span-yes.zip

ls -l span-*
-rw-r--r-- 1 me me 45000 Jul 26 14:41 span-no.txt
-rw-r--r-- 1 me me   349 Jul 26 14:42 span-no.zip
-rw-r--r-- 1 me me 58000 Jul 26 14:41 span-yes.txt
-rw-r--r-- 1 me me   397 Jul 26 14:42 span-yes.zip

Actually, every other line should be some unique string, but this should give you an idea of what is going on when the characters before and after <span></span> are often the same.