Directory sanity

[Robert Olson]

I'd like to get a consensus on the overall directory structure for the AG 
software as installed, as I'm getting quite confused trying to understand 
the way things are now. This will also go toward making the path clear for 
how to fully run the system from a CVS checkout.

I see the following classes of information that need to be stored (this 
summarizes and repeats my earlier mail on the topic).

Per-user configuration. Client profiles, identity certificates, 
user-specific trusted CA certificates, various per-user preferences.

Per-system configuration. Systemwide trusted CA certs, node service configs

Application code. The actual AG software, client code, server code, and 
supporting libraries.

Service code. Services that are packaged up to be sent out as needed.

Downloaded service code. Local copies of services that are actually executed.

Service configuration information. Is this separate from per-system 
configuration? Is there an issue of file-writability? That is, is only the 
process that runs the services going to write to this information?

Am I missing anything? The next step is to assign each of these classes to 
a standard directory on the system, and to define exactly what discovery 
mechanisms there are for finding the correct information. My 3/7/03 mail 
discusses a first pass at such an assignment. 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/ag-dev/attachments/20030321/a4e2e4d6/attachment.htm>