Brock Pytlik wrote:
Tom Mueller wrote:
Brock Pytlik wrote:
There are three distinct entities, publishers, streams, and
repositories. Publishers are the entities who put together
distributions, sign packages, etc... Specifically, they distribute
one or more streams (or trains or whatever term we settle on. dev
and release are two examples of streams for the opensolaris.org
publisher.). Repositories are simply collections of packages,
possibly from multiple publishers and multiple streams from those
publishers.
About repositories - first, I'm assuming that there is a distinction
between a pkg.depotd process and a repository. A repository would be
identified by a unique URL, but a pkg.depotd process might eventually
service multiple repositories. Is that right?
No, I think we're suggesting moving to a model where there's a
one-to-one mapping between pkg.depotd process and a repository. But
that depo/repo/process might serve up packages from many streams and
many publishers.
Essentially, we want to move to a model where a repo/depo is just a
container that hands out bits of packages to clients as needed. In
short, the average user should never need to care what repo(s) they're
connected to.
Does that make things clearer?
Not yet. Shawn's reply said that a depot would eventually host multiple
repositories. This is requested in bug 7653.
How many type of URIs are there that a user needs to know about. Some
of the responses have talked about URIs for a publisher. What would
such a URL look like? For example, for OpenSolaris.org, would the
publisher URI be http://pkg.opensolaris.org/. And when this is
accessed, do you get a p5i file that has the list of streams and the
URIs for the repositories for those streams? For example,
http://pkg.opensolaris.org/release and http://pkg.opensolaris.org/dev.
Are those URIs for streams or repositories? If a repository is hosting
many streams and/or many publishers, what are the various URIs for those?
You didn't address the issue below about the distinction between a
stream and a repository.
I'm trying to apply this to the SWI case where we having community and
Sun editions of various different products such as GlassFish, WebSpace
Server, Webstack, MQ, MySQL, and OpenDS. In this case, the publishers
might be Sun (for all Sun-branded versions of those), and then
pkg.<project>.org where <project> is different for each one. We would
have dev, release, support, and contrib streams for each product, so for
the Sun publisher, we would have 6*4=24 streams. Then we have 6
OS-platform-specific repositories for each one. So we have to have 144
repositories and depotd processes and each repository has its own URL.
However, through the use of the User-Agent header and an Apache
front-end, users typically only see URLs for the 24 streams.
I'm not yet sure if having the concept of a "stream" makes this easier
to explain than just talking about "repositories".
Thanks.
Tom
Brock
So assuming that a repository is identified by a unique URL (possible
with mirrors that are identified by other URLs), what would be an
example of where multiple streams would be served from a single
repository? Also, what would be an example of where packages from
multiple publishers would be coming from a single repository?
I'm wondering if the world would be simpler if we just chose to limit
a repository to containing packages from a single stream from a
single repository. This would make a 1:1 relationship between stream
and repository - effectively the two concepts are the same.
Thanks.
Tom
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
