revision-history considered harmful

[John A Meinel]

Robert Collins wrote:
> There was a discussion late at UBZ about removing revision-history as it
> appears to be somewhat harmful..
> 
> We've had a number of users use the property of pull to traverse any
> link result in revnos vastly different than what they considered
> 'upstream'.
> 
> The revision-history file is somewhat special - its not reconstructable,
> which makes it precious, its only identified by a URL, which makes URL's
> important (where all our other data is not sensitive to things moving
> around).
> 
> Martin and I sketched an approach where commits record the revno they
> had at the time of commit, which combined with the branch nickname means
> that doing a 'switch' of the revision-history when you 'pull' would be
> understandable to users: you would *always* end up with the revno of the
> branch you pulled from. 
> 
> We thought we could trial this by simply making the revision-history
> replacement function of '--overwrite' always be on, without disabling
> the divergence check. If that satisfied people, we can nuke
> revision-history-the-file.
> 
> Rob
> 

I'm +1 the concept of always writing the complete ancestry. Basically
getting rid of convergence.

I don't know if we want to actually delete the revision-history file, as
it seems to be a decent optimization for stuff like "-r 12".
But yes, I think we could change pull.

John
=:->


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20051222/5f2313df/attachment.pgp