On Wed, Sep 03, 2008 at 07:04:05AM -0700, Leslie Wu wrote:
> While it's great that Processing.js and the like exist, to be really bold
> and experimental, we (in the advanced prototyping / research / hacking
> space) need to dip and dive deeper than modern notsofreaky sandboxes allow,
> and hope that HacketyHackers won't be confined to such small but Flashy
> spaces.

On the note of sandboxing, I think Chrome has made me ditch my prior
ideas about how I want to handle security in Shoes.  (Because that's
the big problem here, not a bunch of little platform bugs.)  I don't
think I'm going to use the sandbox extension after all.  It's too
experimental for right now and not generic enough should I end up
supporting Rubinius of Ruby 1.9 or anything else.

I think Chrome (and Yahoo's BrowserPlus) have a fundamentally nice
idea: keep the interpreter in its own process and then sandbox the
process.  I'm thinking something like a cross-platform Sandboxie
(see: sandboxie.com) which emulates a filesystem and socket
environment for each process.  This keeps it nice and generic and
doesn't require a complex API.  The API would be much like the
popen3 extension, basically.  With a config describing which services
to cut/redirect, I guess.

I still think Shoes is best suited to compete with GUI toolkits
(since they are just so lousy) rather than to take on Flash.  But
once Shoes is solid, I hope to get cracking on the security stuff
and, well, who knows what's in store after that.

_why

Reply via email to