On 11/29/11 5:06 PM, Rob Weir wrote:
On Tue, Nov 29, 2011 at 8:51 AM, Gavin McDonald<[email protected]> wrote:
-----Original Message-----
From: Rob Weir [mailto:[email protected]]
Sent: Tuesday, 29 November 2011 10:15 PM
To: [email protected]
Subject: Re: Business models that will not work [was: Re: Can we update our
migration status table?]
On Mon, Nov 28, 2011 at 10:09 PM, Pedro Giffuni<[email protected]> wrote:
FWIW;
I think a central site for hosting AOO extensions would be welcome. It
would be fine to have such a site sponsored by donations and web
publicity, and offering a share of technical support and commercial
extensions would be fine too.
Compare, for example, Apache Maven with Sonatype's Maven Central.
This is similar to AOO and Extensions.
I haven’t really read this thread, but I have a server here in AU if there
really is
nowhere else to host this stuff. Are there any specs on what's required?
Or, shoot me if I got the wrong end of the stick.
Yes, that is the wrong end of the stick. But it is an interesting
side of the stick as well, so I'll move it to its own thread, so we
can explore the options.
We currently have an extensions repository and a template repository,
kindly hosted by OSUOSL (http://osuosl.org/) on our subdomains:
extensions.services.openoffice.org
templates.services.openoffice.org
These repositories carry 3rd extensions and templates under a wide
variety of licenses, including a mix of open source licenses, copyleft
and non-copyleft, as well as non-open source licensed "freeware" and
"trial-ware" packages. Because of these various licenses, I think it
is unlikely to be something that we could host at Apache, even in just
binary form.
So what are our options?
==Option 1: Remain at OSUOSL==
We could remain with OSUOSL hosting. However, the existing site is
very unstable. For this approach to be practical we'd need a
volunteer with Drupal skills to work with OSUOSL to diagnose what is
wrong and to restore stability to these services. Maybe some sight
maintenance, upgrades or tuning is sufficient?
I think this would be the ideal short-term solution at the very least.
==Option 2: Move Critical extensions to stable host==
This is more a back up plan if nothing else works in the AOO 3.4
timeframe. There are a handful of critical extensions that many users
will want access to, such as spell checking dictionaries. If OSUOSL
is not stable when 3.4 is released, then we will have many thousands
of very frustrated users. So with this option we copy the critical
extensions to Apache-Extras and point 3.4 users to that.
==Option 3: Clone OSUOSL repositories to another host==
In this option we rehost the existing repositories at another host.
So similar to Option 1, we would need a volunteer with Drupal
expertise.
==Option 4: Host repositories elsewhere, using new UI==
SourceForge, etc. There are ways, for example, to create
meta-communities ("Neighborhoods") at SourceForge across projects.
Similar things can be done at Google Code or elsewhere. Take
advantage of these forges rather than maintaining our own customized
app.
==Option 5: Re-architect the Repositories==
This is the option I personally favor for long term. The downside of
what we have today is we have a single repository with a single host.
This is non-optimal for several reasons. There are different needs
out there if we look at downstream consumers and/or enterprise users.
They may want to have a restricted or supplemental repository. I may
want to make available to my users an interface that shows only
non-proprietary extensions, plus non copyleft templates plus my own
proprietary house templates. Someone else might want to restrict
things differently. Look at analogous things with Ubuntu component
repositories, with the ability to disable or enable proprietary
drivers and/or non-supported components.
In a new design we could start up front with a specification for a
data model for an extension, something expressed in simple XML, with a
query/fetch interface as a RESTful service. This would allow multiple
repositories to look and behave identically from the data perspective.
That then allows websites that aggregate such repository data, as
well as in-app browsers for extensions and templates. Then you can
imagine AOO allowing a user or admin to enable any from a list of
several extension repositories to work with.
The other thing this approach does is separate the extension metadata
from the actual licensed extension. If we wanted to have a canonical
repository of registered extensions, but without actually hosting or
storing the extensions, then that should be OK. We're hosting URL's
to resources. We're not distributing code.
(There might be some hybrid way of doing this as well, by using
SourceForge or similar for the underlying hosting, but then with tags
or a common extension.xml in the root of each repository, then
publishing data that we create our catalog from)
today we have one default repository but each extensions can carry it's
own update Url. That means in detail
- no update Url in the extension package, the default Url is used to
search for update information for this extension
- update Url in the package, this one is used to search for update
information.
Such an update Url can link to a simple xml file providing the
extensions meta data about available updates. This file can of course
provide information about more than one extension.
Based on this we can probably improve the whole extension management a
lot. And with some extra work we can provide a filtering and configuring
mechanism that you probably have in mind.
We should keep in mind that we do also some virus check (as far as i
know) for the extensions hosted on the OSUOSL repo.
Independent which way we prefer in the future, but we should ensure that
we have some control over the submitted extensions. I don't know how
exactly this can be ensured. I think about something like a review
process, or certifying. Simply to ensure that at least in an official
repo we don't have any malware.
Juergen
(Happy to do the migration too if that’s needed)
Gav...
