Doctypes and charsets.

[Stodder]

Hi guys,

I promise not to ask another question here for the next few days, but I just had to get an opinion on this :

What's the best doctype and charset to use for creating Kindle-compliant mobi files? I'm thinking of using XHTML Strict and ISO 8859-1.

Will this be compliant, and will it be able to handle named character entities (— et al)? My validation has come up clean, but I'm never sure what to make of charsets and the associated dtd stuff.

Any help would be great :-D

Stodder

[pdurrant]

Originally Posted by Stodder

Hi guys,

I promise not to ask another question here for the next few days, but I just had to get an opinion on this :

What's the best doctype and charset to use for creating Kindle-compliant mobi files? I'm thinking of using XHTML Strict and ISO 8859-1.

Will this be compliant, and will it be able to handle named character entities (— et al)? My validation has come up clean, but I'm never sure what to make of charsets and the associated dtd stuff.

Any help would be great :-D

Stodder

I'd recommend XHTML 1.1 Transitional, but be sure to limit use to XHTML that's valid in the ePub 2.0.1 spec.

And then you'll probably want to limit even more to things that convert to Mobipocket well using kindlegen.

You should definitely use utf-8 as the character encoding, as that's a requirement of the ePub spec (well, or UTF-16, but that would just be silly.), and is also handled well by kindlegen.

[Stodder]

@pdurrant: Thanks for the detailed answer Yes, it's probably better for me to go with transitional rather than strict XHTML.

However, I thought UTF-8 was not workable in Kindle (Amazon: Formatting Your Book > Character Encoding and Supported Characters)?

This is soo complicated o_O!

[Stodder]

Ooh, another question: if I use ISO 8859-1 for the source html (and I have no idea if that's a good idea, anyway), do I choose Western 1252 in the Mobipocket Creator encoding settings for the output MOBI file? The program only has 2 choices: UTF8 and Western 1252.

[pdurrant]

Originally Posted by Stodder

However, I thought UTF-8 was not workable in Kindle (Amazon: Formatting Your Book > Character Encoding and Supported Characters)?

That page is talking about Amazon Kindle Direct Publishing, where you send them an HTML file and they convert it to a Kindle file. You don't want to do that. Use the free kindlegen locally - it'll be faster and give you more control.

Amazon Kindle ebooks most certainly can use utf-8. I have several published at the moment that use it.

[pdurrant]

Originally Posted by Stodder

Ooh, another question: if I use ISO 8859-1 for the source html (and I have no idea if that's a good idea, anyway), do I choose Western 1252 in the Mobipocket Creator encoding settings for the output MOBI file? The program only has 2 choices: UTF8 and Western 1252.

If you were to use ISO 8859-1, you should use Western 1252.

But I'd recommend utf-8. And while Mobipocket Creator is OK, I'd recommend using kindlegen for your final uploads to Amazon. (Mobipocket Creator hasn't been updated for a long time, and doesn't permit graphics bigger than 63KB, and probably has other limitations.)

[Jellby]

The safest is just to use ASCII, and convert all other characters to numerical or named entities:

á -> á
ü -> ü
— -> —
’ -> &rsquo (or & #8217;)

If you do this, you won't have to worry about encodings, because the HTML text is actually just plain ASCII.

[Stodder]

@pdurrant: ah, I see now

Hmm, maybe I should give kindlegen another go. I was not crazy at first, but it could be for the best to branch out from MBPC. I couldn't find a good manual to kindlegen, and that also discouraged me.

@jellby: never thought of just using ASCII. All my special characters have been replaced with named character entities already.

[Stodder]

Oh, thanks for getting me onto KindleGen, Pdurrant!

I've started using it and its really useful. I created the html as XHTML Transitional Charset UTF8 and put in named character entities. The KindleGen output seems to read nicely, so hopefully the characters don't prove troublesome in other readers?

A part of me keeps thinking: use numeric entities. But I read once that older kindles have trouble with numeric entities--don't know how true this is?

[pdurrant]

Originally Posted by Stodder

Oh, thanks for getting me onto KindleGen, Pdurrant!

I've started using it and its really useful. I created the html as XHTML Transitional Charset UTF8 and put in named character entities. The KindleGen output seems to read nicely, so hopefully the characters don't prove troublesome in other readers?

A part of me keeps thinking: use numeric entities. But I read once that older kindles have trouble with numeric entities--don't know how true this is?

I suspect that Kindlegen will convert numeric entries into utf-8. But I haven't looked at the output to see. If you get hold of MobiUnpack.py you can test it yourself. (It's available here on mobileread somewhere.)

[Stodder]

^ I've poked around in the mobi files I made with Kindlegen, and they seem to retain named entities, but never thought to check whether they converted numerical ones.

But thinking about it, I didn't chekc if the output MOBI was UTF.