Yoric at users.sf.net
Tue Aug 7 15:08:07 EDT 2007
Okay, I surrender and return to the oblivion from which I was spawned.
I've obviously been working with quite different assumptions about what
was easy and what was hard (in this case, getting cooperation both from
the server and from the OS). While most cscott's arguments are sound in
the context of OLPC, unfortunately, relatively little may be applied in
the context of a client-side, multi-format, cross-platform,
mostly-disconnected, easy-to-extend e-Book reader software such as
So, all the best and good luck for the future.
On Tue, 2007-08-07 at 12:43 -0400, C. Scott Ananian wrote:
> On 8/6/07, Ian Bicking <ianb at colorstudy.com> wrote:
> > If we want to load up these persistent browser caches using a non-HTTP
> > connection (e.g., carrying them around on USB drives), then we'll need a
> > file or archive format for that, something that includes both metadata
> > like location, ETag, content-type, etc., and the actual file bodies.
> Right. We need a bundle format. I believe sj et al have already
> written something like this for Library data? The whole 'stick a USB
> key in and transparently get content added' mechanism is getting to be
> pervasive throughout the XO interface. I don't really see any
> difference between magically discovering a (say)
> alice-in-wonderland.zip file on a USB stick and magically acquiring a
> link somehow to http://canonical.source/alice-in-wonderland.zip --
> both trigger a download of the complete book. In both cases the
> archive format should have enough metadata to indicate its source (and
> for some applications, its authenticity). This gets shoved into the
> library cache.
> Again, the big picture here is that the whole scheme works even if we
> *don't* have all these magic pieces. The magic each does specific
> things: download a bundle at once from a URL, install resources from a
> key, etc. We don't have to reinvent the pieces which already work:
> like browsing individual pages from the web and bookmarking them.
> Finally: the working assumption for the XO project (repeated in many
> places) is that the machines usually *do* spend most of their time
> on-line. That is the point of the pervasive mesh networking and etc.
More information about the Devel