One repository or many?

[Julian Green (KMIS)]

I know that there is no absolute answer to this question but our Perforce repository structure is:-

//common/...
//clients/client1/...
//clients/client2/...
//clients/client3/... etc.

In JIRA we have one project per client. Each project connects to the common repository and a single client repository.

What are the advantages/disadvantages of:
a) Having a minimal fisheye repository
ie have 2 repositories in fisheye (//common/... and //clients/...) and connecting all JIRA projects to both
and defining a repository path (of /clientx) for each client.
Would defining the project repository path in jira inpact the reporting of changes to /common/...?

or

b) Having a more comlex repsitory structure in fisheye
ie defining separate repositories for each client (//clients/client1/... , //clients/client2/... etc)
in this case the JIRA cnfiguration is simplified as there is no need to specify a path for each project.

There seems to be a trade-off between a complex setup for FishEya and simple one for JIRA and vice verse.

I am not asking which is better for me, that has to be my decision, but I am not clear about the impact (advantages/disadvantages) of these options.

[Norman Abramovitz · Accepted solution]

I may be wrong here because I am still evaluating Fisheye, but I preceive FishEye like a virtualized source control system. So you can get a consistent view across multiple source control respositories. FishEye also seems to provide a more integrated feel with Jira in comparison to the subversion plugin as an example.

So I would based my decision based upon security issues verses the administrator ease of use. For example, can everyone see the repository changes?

From what I can tell your end user experience would be the same in either scheme. So if the simplier design works, I would suggest going with it.

[Doug Bass]

Having permissions on the fisheye repository is one consideration. The other consideration is threading. Two fisheye repositories can be scanned in parallel if fisheye has enough scanning threads configured. This would be useful if one perforce branch had lots of changes and the other branch had fewer changes.