[ANN] Monticello Versioning

Colin Putney cputney at wiresong.ca
Sat Jul 26 08:11:56 UTC 2003


On Saturday, July 26, 2003, at 12:22  AM, Daniel Vainsencher wrote:

> The word overhead is being used to generically here for me. I'm not 
> sure
> what issue Colin is talking about. Loading Monticello? What overhead 
> are
> you talking about - overhead for the developer to use two formats?

By overhead, I meant that you have load Monticello in order to load an 
mcv file into your image. Ok, it's not *that* big a deal, but it seems 
to introduce a needless dependency.

> I think that using two formats adds overhead for users to become
> patchers, which Colins "send note with history info" solves only
> partially. Tool support could make getting the mcv trivial, if the st
> packages come with a url, which could almost eliminate this difference.
> One would simply have an option for "to work on this package, load
> monticello, get the mcv, and start working". It would, however, be 
> quite
> foreign to everybody (have a hard time think where one could put this
> option, even).

Well, no. The point of my proposal is that users can become patchers 
just by installing Monticello. They don't have to go get an mcv, 
because they already have both the code and the version history in 
their image.

> Giving the user an mcv is simpler. The current problem with it is that
> lacking an SMMCInstaller, one's package is hidden by default because it
> is not auto-installable (so it might take a while for package users to
> even notice - the problem is invisible). How about a trivial
> SMMCInstaller that loads MC if it is missing and then loads the mcv?

Yup, that sounds good. It solves the problem at hand, and is much 
simpler than my proposal.

Colin



More information about the Squeak-dev mailing list