On Wed, May 12, 2010 at 10:50:07PM -0500, Shawn Walker wrote:
> On 05/12/10 09:01 PM, johan...@opensolaris.org wrote:
> >     http://muskoka.eng/~johansen/webrev-pkg-mdns-2/
> 
> modules/server/depot.py:
>   line 107: Upon further reflection, I'd like to not automatically
>       disable the publisher operation just because mirror mode is
>       active.  Specifically, I'd like to see it omitted from the
>       list of operations in __init__ based on whether the publisher
>       prefix property is set for the cfg_cache.

I'm not sure I understand what case you're trying to address here.  A
mirror can serve content for multiple publishers.

Orthogonally, there's another issue with how this data is currently
generated and distributed.  I took a look at the data from the
publisher/0 for mirrors on SWAN.  The majority of these mirrors aren't
returning any correct configuration information.  My sense is that
there's a RFE lurking here for better configuration of publisher/0
information if none is specified when the depot is set up.

Allowing the client to use the mirror as a source of configuration
information when we may not know all of the different origins of the
content, and know the data is frequently wrong seems troublesome.

In your on-disk format proposal, you claim that we'll always know the
source of the content that's stored on-disk.  I'm not sure how this will
work for existing mirrors, but for new mirrors we could easily
auto-generate the publisher prefixes by looking at what repositories
we're serving.  This seems like a more roubust long-term solution.

> modules/server/repositoryconfig.py:
>   lines 61-62, 71-72, 82-83: nit: just return instead of assigning to s
>       as is done on lines 91, 98, etc.?

Fixed, thanks.

> pkgdefs/Makefile:
>   line 21: copyright needs update

This file actually doesn't need to be in the changeset anymore.  Let me
see about removing this.

> tests/
>   Possible to add a simple unit test of some kind?

What did you have in mind?  The mDNS stuff is disabled by default, and
we don't currently distribute PyBonjour.  The test client and server
must be running the mdns daemon in order to publish and recieve the
advertisement.  I've tested this by hand, but I'm not sure what we could
test that's easily automated.

Thanks for taking another look at this.

-j
_______________________________________________
pkg-discuss mailing list
pkg-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to