updatinator benchmarking (Was: rsync benchmarking)

[Ivan Krstić]

Alex,

On Jul 10, 2007, at 10:21 AM, Alexander Larsson wrote:
> I've been doing some more work on updatinator.

I'm happy to see this work continuing, but I want to set clear  
expectations for FRS. I will not give security signoff for the FRS  
update system, in my mind one of the most critical subsystems on the  
machine, to be based on new code.

I want to explore other update mechanisms after we ship, and  
updatinator tops the list. I don't want to discourage your work. But  
given where we are with our software and where we need to be a couple  
of months from now, we don't have the in-house resources to give  
updatinator the kind of scrutiny that would put my professional  
paranoia at ease. We take various software risks at different places  
in our stack precisely because we're relying on the update system to  
work flawlessly. With existing, well-known tools, I can make  
reasonable guarantees about that being the case. With new ones,  
despite your obvious talent and attention to this problem, I can't.

An rsync-based solution, possibly without even Scott's wrapping  
manifest code, is the only update approach I will support for FRS.  
After that, we can revisit the issue.

Cheers,

--
Ivan Krstić <krstic at solarsail.hcs.harvard.edu> | http://radian.org