[cap-talk] Tubes: Capability-based Sharing By Any Other Name?
toby.murray at comlab.ox.ac.uk
Tue Jan 16 04:35:36 CST 2007
I just came across this application, Tubes, from www.tubesnow.com .
It's a Windows file-sharing client that appears at first glance to apply
a very capability-like interface to file sharing. The interface it
presents to users for sharing files seems scarily similar to the
"capability directory" / "gift directory" approach that's been advocated
so often on this list.
The directories here are called "tubes". A user can create a directory
and then invite other users to access that directory. Invitations are
sent by email. When inviting a user, one is able to determine the sort
of access that user will have to the directory: read-only, read/write
The application implements a model where the contents of a shared
directory (a tube) is cached (replicated) on all of the machines of uses
who have access to the directory.
Any user with read access to the tube automatically has the tube's
contents cached locally on their machine. Modifications to the local
cache by users with write access automatically propagate to other users
with read access. (At least, that's how my brain has interpreted the
company's high-level descriptions of their product).
I have no idea if some central server keeps a master copy of each
directory or not but I suspect it might be necessary in order to have
this model work for users who don't leave their machines on-line very
I haven't yet pulled down the application but am keen to get a good look
at the UI. There may well be something here to learn about UI design for
this sort of sharing paradigm. Hopefully they've come up with some neat
ways of informing user's about their security choices and keeping the
user informed about past security-relevant decisions. Particularly, in
regards to the authority they have given to others to access the
directories that they have created.
It'll be interesting to see how this paradigm flies in a widespread
To E people: this looks to me like one of those killer apps that would
be awesome to see implemented in E, since E seems perfectly suited to
implementing something like this as an existence proof of the power of
the capability paradigm. Are their any features currently missing from E
that would preclude the open-source development of something like this?
More information about the cap-talk