Shawn Walker wrote:
On 08/17/12 13:38, Philip Brown wrote:
Given that there is an expected(?)separate step of "pkgrepo refresh", it
seems like the pkgsend stuff could be made to stay solely under the
relevant publisher subdir, and postpone the other stuff for the refresh
step?
Currently, writeable access to the root of the repository is required
for all clients that want to manipulate it.
>
....
Clients manipulate a repository through a "generic" internal API. Put
differently, the same publication and administration logic is shared by
all clients and no special logic exists for specific clients, so the
same restrictions exist for all of them for the same sets of operations.
I was going to ask what files outside of the publisher area were being
touched, but then decided to just look for myself, with truss.
It appears that there are some serious inefficiencies going on here.
This was a test of doing a pkgsend publish, of one specific package, for one
specific publisher. A single command line "pkgsend publish blahblah..."
Okay, it touches some top level things, such as pkg5.repository
And okay, it touches a bunch of files under the publisher specific area,
such as files under /index/ and /catalog/
But then it goes and looks at the same files under the index, and catalog,
directories, of EVERY OTHER PUBLISHER
This is very wrong :(
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
