Experimental Physics and
Industrial Control System

"J. Lewis Muir" <[email protected]> · Thu, 8 Feb 2018 12:14:59 -0600

On 02/07, Mark Rivers wrote:
> On 02/07, Tim Mooney wrote:
>> I like the certainty of using tar files of released versions in most
>> cases, and explicitly deployed git clones in a beamline's local file
>> tree if it's required.
> 
> I think "git checkout R3-4" is as certain as getting the R3-4 tar
> file.
> 
> The advantage of the git clone is that you can immediately see if
> someone has changed a file with "git diff R3-4".  If you have a tar
> file, and someone has changed a .template file, etc. you have no easy
> way tell.

I tend to agree with Mark.  Still, I think it wouldn't be too hard to
support *both* approaches: one fetches source from an archive (e.g.,
.tar.gz) while the other fetches source from a Git repository (or other
supported VCS).  (For the fetch from an archive, it would be nice if
there was a manifest that included a checksum (e.g., SHA256 hash) for
each archive downloaded to ensure the archive wasn't changed.  Git has
this built in.)  The archive approach would probably define a minimal
feature set, and the Git approach would be able to implement that.
The Git approach could then provide more features beyond the archive
approach if it wanted to, or it could just be that the result is a bunch
of Git repositories, and anyone who knows how can take advantage of
that.

Lewis

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System