Re: [Monotone-devel] database compaction

[Markus Schiltknecht]

Hi,

Thomas Moschny wrote:
> That was motivated by the fact that heights are only there for optimization 
> reasons. They are caching the topological ordering of all revisions present 
> in the database and thus don't represent first-order information. The heights  
> can be thrown away and regenerated at any time (like rosters). So, we wanted 
> to keep that information separately.

Understood.  I'd argue that there's not much difference between an extra 
column and an extra table, regarding safety of data. So this distinction 
of first vs. second-order information is just a structural visibility 
issue, not a real separation into more and less important information.

Regards

Markus