Hi Fabrizio and Andy,
In order to test the posix-like client in the situation where we initially
intend to use it, could we make an application "xcp" or "xrdcp"? (I don't care
strongly which name we use. "xcp" is shorter, of course, but it looks like
there are already things out there called "xcp". Does either of you have a
preference?)
We will want to do things like:
xcp aaa.root root://host:port//some/path/bbb.root
xcp root://host:port//some/path/aaa.root bbb.root
xcp root://host1:port1//somepath1/aaa.root root://host2:port2//somepath2/bbb.root
Andy, I assume you have thought about this. How far do you want to go with
it? Could/should we support other aspects of the semantics of "cp" , i.e.
o wildcards for local files:
xcp *.root root://host:port//some/path/
Obviously the "xcp" command can just disentangle this into the full set
of separate copies and do them one-by-one, hence nothing special is
required from the server, even in a load balanced environment.
o wildcards for server side files:
xcp "root://host:port//some/path/*" .
I assume you don't want to support this due to the dynamic nature of
the catalog when using the redirector. (FWIW, I agree with that.) But
does it make sense if done on a dataserver?
How is the directory treated in a load balanced/cache filesystem environment
for something like the following?
xcp aaa.root root://host:port//some/path/bbb.root
Is "/some/path" just created automatically?
Are there other important subtleties to this, in particular with the load
balancer?
Pete
-------------------------------------------------------------------------
Peter Elmer E-mail: [log in to unmask] Phone: +41 (22) 767-4644
Address: CERN Division PPE, Bat. 32 2C-14, CH-1211 Geneva 23, Switzerland
-------------------------------------------------------------------------
