downloaded files and the Journal

[Tomeu Vizoso]

On Sat, Jun 14, 2008 at 7:36 AM, S Page <info at skierpage.com> wrote:
> I was exploring the OLPC Library in Browse in build 703.  The transition
> from browsing to viewing a PDF in Read is pretty clunky, I noted the
> glitches in
> <http://wiki.laptop.org/go/Talk:Read#Many_issues_going_from_Browse_to_PDFs>
>
> The oddest thing is that every time you click on a PDF, another copy of
> the PDF is saved to the Journal's datastore.
>
> This sounds intentional as part of the fix for "Read doesn't save the
> PDFs it opens (Trac #6729)".  I think the Browse activity is putting the
> file in the Journal, as the extra copy appears in the Journal's
> datastore even before I run Read.
>
> But this doesn't make sense for read-only local files in the OLPC
> library like images > world maps > click Africa.  I've now got 7 copies
> of this .5MB PDF in ~/.sugar/default/datastore/store, each identical to
> /usr/share/activities/worldfactbook-maps/africa.pdf ; seems a
> significant waste of space.  (Or does JFFS2 magically optimize multiple
> copies of the same bits?  df -k reports mtd0 dropping by about the size
> of the PDF.)
>
> If Evince worked as a browser plug-in, maybe the duplicated files would
> go away (along with some other glitches).  However, "Consider unifying
> Read and Web (Trac #3212)" was rejected.
>
> If viewing PDFs remains a separate activity from Browse, then maybe the
> Browse activity could detect that it's "downloading" a local read-only
> file, and rather than the confusing download step leading to a copy in
> the Journal's datastore, it could just pass the local file:/// URL to
> Read on the command line.
>
> But even if Browse bypassed the download step for local files, the
> problem of multiple copies remains for remote files.  I think each time
> I return to a PDF on the Internet, the Browse download step likewise
> makes another copy of the PDF in the Journal's datastore.
>
> Has anyone considered making the Journal a two-level store, so that if I
> download the same file 7 times there's only one copy in the datastore,
> using symlinks or references or other magic?  Maybe this could even be
> made visible in the UI:
> "You've downloaded this file 7 times already, please check the Journal
> before repeating yourself.  Sheesh kid do you think the Intertubes is
> free?!"
>
> Meanwhile, the OLPC Library should prefer HTML or SVG content over
> single-page PDFs.
>
> (The problem of local copies of Web content isn't unique to Sugar.  On
> my desktop PC I'll save stuff from the browser locally then forget that
> I've done so.  I want a browser and O.S. that opportunistically treats
> everything on my hard drive as a local cache of something from the
> Internet that might be updated, or have disappeared.)
>
>
> If I'm wasting time here with obvious rehash, I apologize and cut me
> off.  And/or I can file enhancement bugs for these ideas.

These problems are known of, and hasn't been solved yet because of
lack of man power. If you wanted to search in trac and review that the
tickets in there contain all the relevant information, would be very
nice.

Thanks,

Tomeu