Re: [Wigwam-devel] wigwam presentation material?

[Michael R. <mr...@bl...>]

Jason Stone writes:
> >>I seem to remember fess or paulie asking me about how to make a safe 
> >>setuid chroot wrapper.  did that ever happen?
> >
> >no we didn't get anywhere with that.  That was an attempt to further 
> >reduce dependancies on the base system, another wigwam deficiency is 
> >that the boundry between wigwam and the base system is blurry to say the 
> >least.
> 
> yes, this is, for my money, the biggest flaw in wigwam.  it's all well and 
> good to say that wigwam frees you from local system dependencies, but it's 
> way too easy to start relying on system libraries and not even realizing 
> it - this has bitten me several times.  chrooting is the only way to be 
> sure.  plus, it buys you paranoia points.

We started down the chroot path at Snap last year, but there were
several issues that didn't seem to have a clean solution.  For instance, 
LOCAL_VAR...ideally you could have several versioned chroot playpens
side by side, and you could start any one of them using the same
LOCAL_VAR so that persistent data (e.g. mysql databases) is saved.  We
couldn't figure out a way to do that without hard-linking directories or 
having serving the directory via NFS over localhost so it can be mounted
in several places (if that's even kosher).  Did anyone else ever think
of a better way to do this?

Mike

P.S. I'd say the biggest flaw in Wigwam 3 is that 'packagectl
update-packages' doesn't rebuild packages when their dependencies are
upgraded (whereas 'packagectl upgrade PACKAGE' does).  But that should
be fixable. :)