version.Current must die

[David Cheney]

I agree (although Tim and I have discussed this at length).

My proposal at the time was to delete version.Current, then audit all 
the points that break. My suspicion is they will break down into 
categories like:

* don't need a version.Version, but a series or an arch
* don't need a version.Version, but tools url.
* etc

It might be possible to work backward from there to a less aggressive 
set of changes.

Dave

On 22/03/13 08:30, Tim Penhey wrote:
> Hi there,
>
> Subject line says it all.
>
> version.Current leads to the standard "it works on my machine".
>
> version.Current represents the current release and architecture of the
> machine that the code is running on, and probably not what we want at all.
>
> I propose we have a version.Default() that returns a new instance of a
> version.Binary that represents the current supported LTS (currently
> "precise") on a 64bit OS ("amd64").
>
> I have a branch where I have started this work, but it needs some more
> time to think around the upgrade code.
>
> Even just changing the values in version.Current to "precise" and
> "amd64" causes all tests to pass except for the one that checks that
> version.Current.Series is the same as `lsb-release`.
>
> By having a defined versions.Default(), we are testing the same code on
> everyone's machine.
>
> Comments?
> Tim
>