[OLPC library] questions about PDFs

[Sayamindu Dasgupta]

On Sun, Oct 5, 2008 at 3:40 PM, S Page <info at skierpage.com> wrote:
> Sayamindu Dasgupta wrote:
>
>> [lots of nifty plans for Read]
>
> That's great and undoubtedly worthwhile for legacy documents, but to quote
> Shakespeare, "I come to bury page representations like PDF, not request
> enhancements unto them".  Nobody is contesting my argument that HTML in a
> browser whups PDF upside the head six ways to Sunday; therefore the aim of
> OLPC library advocates should be to get content as HTML (and encourage the
> advanced canvas, CSS, JavaScript and SVG that Browse/Firefox 3 supports),
> rather than making hundreds of thousands of kids who will never print from
> their XO look at "page representations".
>

Agreed. However, there will always be PDF documents, especially in
languages with complex scripts (mainly because of legacy encoding
issues, very hackish fonts, no Unicode support for complex scripts in
authoring applications, etc). So what I'm trying to do is make sure
that we support PDFs as well as we can here :-).

>> I am wondering if I can use the evince library that we are
>> shipping to implement a small and effiecient mozilla plugin, so that
>> the PDF can be directly viewed from within the Browser itself.
>
> That would indeed solve a lot of download-journal-launch issues, and maybe
> the OOM problem when zooming in Read (trac 7090 and 6002).  As you probably
> know, the evince team's position is that rather than provide a plug-in,
> users should use mozplugger (http://mozplugger.mozdev.org/) to run the
> evince app within the browser.  I tried `su -l; yum install mozplugger; yum
> install evince`, but evince wants to install dozens of other packages.
> I wonder when someone will write a PDF parser in JavaScript rendering to
> <canvas> or Cairo.
>

Right now, I'm cooking up a stripped down PDF viewer which can be
invoked by mozplugger to view PDFs (and maybe save them to Journal -
dunno if that is possible right now). I have a basic pdf viewer
working inside Browse now (with security and everything enabled).
Expect more progress soon.

With respect to the zoom problem, it seems to affect the Image Viewer
activity I wrote recently as well (as a result, I had to restrict zoom
to 400% in there) - if someone can do some profiling to figure out
what is going on, it would be great.

>> Some backends which are experimental and are not
>> enabled in our builds include ones for impress files, dvi files, and
>> whatnot :-).
>
> Wiki pages like http://wiki.laptop.org/go/Image_file_formats say that DjVu
> is the best of these, and Chris Marshal on devel list cared about it.  Every
> time I look at the samples at http://djvu.org/docs/ , either Browse or Read
> crashes unless I remember to zoom out (trac 6223). Interesting, but I would
> defer all such bugs until some deployment cares.
>

I noticed some patches to Evince modifying the page caching mechanism
(right now Evince caches around 5 pages in memory, which is not good).
I'll figure out if those patches are already in our version of Evince
or not.

Thanks,
Sayamindu
-- 
Sayamindu Dasgupta
[http://sayamindu.randomink.org/ramblings]