Shawn Walker wrote:
Stephen Hahn wrote:
* Shawn Walker <[email protected]> [2009-02-17 18:29]:
Stephen Hahn wrote:
* Shawn Walker <[email protected]> [2009-02-17 17:53]:
to check on the types of the various fields in the authority/0
response. There are a couple of fields that I believe need to be
list-valued, where the current implementation would only consider
a single value for the field as relevant.
Does "current implementation" refer to the authority/0 proposal?
No, it refers to what we have today, prior to authority/0. (As an
example, we only allow a single origin server per authority.) Reading
5871, I'm mostly concerned about why authority_name isn't required,
and by the growing set of booleans.
authority_name would be required information for the repository, but
not in a .p5i file. It doesn't make sense to require the
authority_name if you provide the origin URL since we should be able
to retrieve the authority name from the origin URL. The only problem
I can see is that if the origin URL is for a depot server that doesn't
provide the authority/0 operation, then we have no name for the
authority.
Perhaps the right answer is to require the authority name in the .p5i
file, but let any authority/0 response from the server override any of
the single value authority fields provided in the .p5i file.
How will the API handle name collisions?
If user clicks on an .p5i file and the name specified is already in use
on the users system, what do we do, follow the BE approach and add a
numeric to the end of it? Will the API resolve this when it does the
authority/0 query against the depot server?
We intend to present an Add New Repo dialog to the user which will have
the Name field filled in but editable, to allow them to change the name
in such a case. Are we saying the user should not be able to change the
name?
Thanks.
JR
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
