Xavier,
In my usage scenario, any sub-project that needs to "publish" (i.e. a
reusable library module) will need BOTH resolvers. They will use the
HTTP resolver to "retrieve" their dependencies, and the SFTP resolver to
"publish" their artifacts after a build.
Subprojects which don't publish (because their code isn't re-used by
other applications / customers) just "retrieve" from HTTP, they don't
need SFTP at all. (These subprojects are still in SVN, just not in Ivy).
I did this for 2 reasons:
1) Performance of the HTTP resolver is much faster, and some of our
users are a long way away over slow VPN or SSH connections.
2) We can give wide access to everybody (HTTP allows public
reading/browsing) and control updates more tightly.
Dave
Xavier Hanin wrote:
Since only one of the two resolvers is used at a time (if I understand
correctly), I'd rather define only one in two different settings files
(with the same name), and include either one or the other file depending
on the environment. You can do this with a naming convention for the
file name and using a property to define which file to load. But it's
mainly a matter of taste, behavior should be the same.
Xavier
