enhancement: support browsing HTML content within bundles instead of unzipping them

[skierpage]

Using the jar: protocol, Browse (or anything that uses XULRunner) can directly access files in a ZIP file.

So instead of unpacking .xol files, just launch Browse and point it at the bundle in the datastore.
jar:file:///path/to/datastore/UUID_of_XOL!/activity_start_path
This saves a copy and on uncompressed file systems a lot of storage space.

I've tried this on my XO-1 running 8.2 with several collections and the Help.xo activity; once Rainbow is disabled or permissions changed, it works fine, seemingly no content or performance issues.

I think the way to do this the content developer may add a line in the bundle's .info file indicating do_not_unzip. Then sugar-toolkit's contentbundle.py would notice this, not unzip during install, and fabricate a jar: URL to launch the content.

Two caveats:
1. Rainbow security disallows Browse access to ~/.sugar in OLPC releases. To workaround this, chmod .sugar or move /etc/olpc-security out of the way.
2. Non-ASCII characters in the .xo filename or the filenames within it cause problems.

I'll try it on a SoaS to see how it works in later Sugar.

[humitos]

We are not using XULRunner anymore and I guess this is not so trivial to implement in our current implementation of Browse (that uses WebKit).

I'm closing this ticket as "wontfix". Please, reopen it if you consider necessary.