Currently, Replica Selectors in Pegasus ignore all file URL's that are not associated with the compute site. This is necessary, as otherwise the Replica Selector may end up selecting a replica specified with a file URL that is not valid on the compute site.

for e.g lets say a job that is to be executed on site X needs file f.a

In the replica catalog the user has

f.a file:///shared/input/data/f.a pool="X"
f.a file:///shared/input-data/f.a pool="Y"

Thus Replica Selector would never select the URL associated with pool Y

Currently this check helps us in filtering out unwanted file urls that have no meaning on site X

However, there is an exception where if Y=local it maybe possible to access these files using the third party transfers if the user has set so.