Re: database server behavior during network partition

[John Hascall]

> > i couldn't agree with this more.  this is one of the core problems 
> > we've attempted to address as we deploy afs worldwide across our 
> > internal wan. 
> 
> Would the availability of read-only access to database servers without
> quorum alter your statements? 

    This would be incredibly cool, ...

> In other words, are y'all (those of you who are, that is) suggesting
> transparent cross-cell volumes as a mechanism for dealing with this
> database problem?  Which, if it didn't exist, would prompt you to simply
> use a single cell? 

    ... on the other hand the fact that setuid-ignoring is a per-cell
    and not also a per-volume (or at least, per-fileserver) flag still
    bites.

John