Shawn Walker wrote:
> Long-term, this will all be made much clearer, in the meantime, this
> might help:
>
> http://blogs.sun.com/srw/entry/what_s_in_a_name

Well it's good that we recognize there is a problem.  :-)

>From your blog it sounds like someone who wants to publish packages must
have their own domain or subdomain.  Doesn't that exclude a lot of people
who would otherwise have no trouble running a server, and making it
available
at a specific URL?

So you seem to be saying:
A publisher is a domain name (forward or reverse)
A repository is a group of URIs

This doesn't seem to map very clearly onto the implementation of pkg where:
A publisher is a random user-assigned name.  (like "osol-dev")
A repository is understood to be a single specific URI. (like
http://pkg.opensolaris.org/dev)

It seems useful to expand on your blog entry with a more careful
specification
before proceeding with the documentation.

The key items seem to be defining the objects and relationships between:
Publisher, Repository, URL, Client side package database

As a side note, what possible use is a domain name that can be either
forward
or reverse?  It's not possible to automatically map it onto a URI.
How do you know if the URI http://tv.guide.magazine.tv/ belongs to the
publisher "tv.guide.magazine.tv" or the publisher "tv.magazine.guide.tv" ?

What are the benefits of creating the concept of a "publisher"?  If there is
no formalism around registration or identification?  At the level of
packaging
technology it seems that defining the concept of "repository" as all you
need to do.

--chris









_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to