proposal/wishlist for OLPC book viewer (was Re: do the directional pad (d-pad or directional buttons) work on B1 hardware?)

[Brewster Kahle]

The PDF's and the DJVU's on archive.org are similar in structure (MRC) 
and both highly compressed and have OCR.  We have 100k books now, and we 
are scanning at over 12k/month-- let us know what you want scanned!

   There is a text layer (OCR with position information) that allows 
searching and cut-and-paste.

   There is a background layer that is a low rez jpg (I believe).

   There is a foreground layer that is a 1-bit JBIG (I believe) layer 
that is compressed.  This basically holds the black-and-white plane of 
the text.

   Then there is a very lowrez foreground color mask just in case the 
JBIG layer is not black.

I will try to launch xbook again, but in the lastest build it did not 
work for me from our website.



An proposal/wishlist for a OLPC book viewer could be:

   * views PDF's (such as ours, see below) rotated in one-page-up view, 
and not rotated in 2-up view.  fill the screen.

   * operate in tablet mode with the directional keys doing navigation 
from one page to the next and some way to zoom into a page and get to 
search.

   * work in "streaming" mode so that the pdf does not need to be fully 
downloaded to start viewing.  (may need a server plugin like djvu does 
to optimal use like jumping around, but it would be good to work off an 
http 1.1 server as well)

   * allow searching inside the book, and then also searching the full 
library (s).

   * some advanced features for browsing a full book such as Michael 
Hawley's 1-bit-font view on a scrollbar, Annotation, reviews, etc.

   * mechanism to get back to the books metadata if that metadata's url 
is imbedded in the pdf. (helps the users and give attribution to 
contributors).

   * launchable from clicking on a PDF on a website.


Any help towards this or something else would be most appreciated.

Onward!

-brewster
Digital Librarian, Internet Archive


ps.  we sponsored the developement an djvu java viewer but it is not 
very active http://sourceforge.net/projects/javadjvu/ look at our djvu 
links for a live demo.   PDF and javascript viewers (like our flip book) 
have more momentum at the moment.


Bert Freudenberg wrote:
> So it appears it is just scanned bitmaps, you're not doing OCR on them.
> 
> I wonder ... is there an image compression method taking advantage of 
> the subject being printed text?
> 
> - Bert -
> 
> Am Jan 21, 2007 um 8:47  schrieb Brewster Kahle:
> 
>>
>> we can construct anything, but pdf is the best for us-- we have them 
>> built now.
>>
>> the old revision of software used to play our pdf's (but crash after a 
>> minute or so), but they dont run at all in the new release (by 
>> clicking from the web browser).   the "flip book" javascript viewer 
>> works, but the pages should be one-up and rotated so they fill the 
>> screen.
>>
>> Here is an example book:
>> http://www.archive.org/details/flatlandromanceo00abbouoft
>>
>> flip book is here:
>> http://www.openlibrary.org/details/flatlandromanceo00abbouoft
>>
>> pdf is here:
>> http://www.archive.org/download/flatlandromanceo00abbouoft/flatlandromanceo00abbouoft.pdf 
>>
>>
>> thank you for the help.
>>
>> -brewster
>>
>>
>>
>> Jim Gettys wrote:
>>> What format is the book source in?  HTML, PDF, bitmap?
>>>                              - Jim
>>> On Sat, 2007-01-20 at 18:49 -0800, Brewster Kahle wrote:
>>>> thank you for the quick reply.   we will be hacking something tomorrow.
>>>>
>>>> would you guess the quickest way to get a demo is browser/javascript?
>>>>
>>>> the other approach seems to be python GTK.
>>>>
>>>> Any early guidance to keep us from a rathole would be appreciated.
>>>>
>>>> -brewster
>>>>
>>>>
>>>> Jim Gettys wrote:
>>>>> Hi, Brewster,
>>>>>
>>>>> This sort of question would normally go to the devel at laptop.org list,
>>>>> where the low level types hang out.  I might discuss the detailed UI
>>>>> design on sugar at laptop.org, where the UI people tend to hang out, and
>>>>> this list for library related items themselves.  At least that 
>>>>> would be
>>>>> my take on it.
>>>>>
>>>>> Yes, the buttons should work.
>>>>>
>>>>> They appear to be keyboard keys.
>>>>>
>>>>> I don't remember if we have X keysyms all hooked up for them this
>>>>> instant or not, but it should be possible without much trouble: we 
>>>>> only
>>>>> got rotation running very recently.
>>>>>
>>>>> On the BTest-1 systems, holding multiple keys down at once may not
>>>>> result in separate key events; this should be fixed in BTest-2.
>>>>>                                 Best regards,
>>>>>                                       - Jim Gettys
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On Sat, 2007-01-20 at 18:02 -0800, Brewster Kahle wrote:
>>>>>> We are trying to make a quick and dirty demo of a few images of 
>>>>>> book pages showing up on the screen.    tablet mode, we think, 
>>>>>> will make the best demo.   we would like to use the directional 
>>>>>> buttons to flip between pages.
>>>>>>
>>>>>> In preparation for a programmer working on this (sorry if this is 
>>>>>> obvious) does anyone know if these buttons send a character to the 
>>>>>> machine?
>>>>>>
>>>>>> also, is this the right list for talking about book viewers?
>>>>>>
>>>>>> -brewster
>>>>>> Internet Archive
>>>>>>
>>>>>> _______________________________________________
>>>>>> Library mailing list
>>>>>> Library at laptop.org
>>>>>> http://mailman.laptop.org/mailman/listinfo/library
>> _______________________________________________
>> Devel mailing list
>> Devel at laptop.org
>> http://mailman.laptop.org/mailman/listinfo/devel
> 
> 
> 
>