Re: [Monotone-devel] Re: [RFC] versioned policy -- introduction

[Nathaniel Smith]

On Thu, Sep 07, 2006 at 10:57:24AM +0100, Bruce Stephens wrote:
> Am I right in assuming you're intending the latter: that (as now) if
> someone has netsync write permission, then they can transmit arbitrary
> stuff to whatever branches they like; but the process of ignoring it
> will be more automatically handled for everyone in the relevant
> projects?

Basically, yeah.  It might well make sense to have servers take
communal trust into account when deciding what to serve, for the
reasons mentioned downthread.  In some sense, though, netsync
permissions are uninteresting.  Any given transport might carry some
bits or not; you could still get them some other way, potentially, and
we don't want communication to become a trust-laden activity.

(Imagine if you could only have conversations with people you fully
trusted :-).)

> Also, is this one trust seed per database?  So does this force one
> database per project (quite probably a good way to do things, but not
> the way everyone works currently)?

Maybe!  Or maybe not; nothing is set in stone yet.  But a 1-1
database<->trust seed relationship is tempting.  For pretty much any
operation, we need to know what trust seed to use, up-front -- this is
dramatically simplified if there is only ever one trust seed to choose
from.  And one database per project already seems to be the best
practice anyway; I don't generally feel guilty about encouraging
people to use best practices.

Servers shared by multiple projects might well be an exception to
this, though.

-- Nathaniel

-- 
In mathematics, it's not enough to read the words
you have to hear the music

This email may be read aloud.