RE: [Openhpi-devel] Re: [discuss] one resouce in more than two domains

[David M. <dav...@au...>]

In the B.01.01 specification, the domain that is accessed when
SAHPI_UNSPECIFIED_DOMAIN_ID is used in the saHpiSessionOpen() call is
assigned by the implementation, and needn't always be the same domain.
Thus, different users can be given different "starting" domains for
discovery.

Also, there is no reason that users cannot open a specific domain ID using a
priori knowledge of which one they want.  This isn't useful for purely
portable applications, but it might meet some specific needs in particular
environments.

David McKinley
Chief Technology Officer
Augmentix Corporation
9724 Beechnut, Suite 125
Houston, TX 77036
713-271-2700 ext. 115
David.McKinley@Augmentix.com
 

> -----Original Message-----
> From: ope...@li... 
> [mailto:ope...@li...] On Behalf 
> Of Sean Dague
> Sent: Monday, June 28, 2004 8:34 AM
> To: ope...@li...
> Subject: [Openhpi-devel] Re: [discuss] one resouce in more 
> than two domains
> 
> On Sun, Jun 27, 2004 at 11:22:18PM -0500, David McKinley wrote:
> <snip>
> > There is one way, however, that the given configuration can 
> be legal.  
> > If domains do not reference each other at all (either directly, or 
> > indirectly through a "chain" of DRT entries), then there are no 
> > restrictions on what resources may be members of each 
> domain.  This is 
> > allowed so that implementations can be created that provide 
> different 
> > users access to different, completely disjoint sets of 
> domains.  (How 
> > domain access is limited to individual users is implementation 
> > specific.)  So, as long as Domains 1, 2, and 3 do not contain DRT 
> > entries that reference each other, there is no problem with 
> the resource membership as described.
> 
> This seems odd, as how would you ever find these domains?  My 
> view of HPI is that the discovery process always starts with 
> a single session opened on UNSPECIFIED_DOMAIN.  Every other 
> domain must come from connections of that Domain through the DRT.
> 
> Are you suggesting that it is valid for an HPI implementation 
> to have hard coded domains not connected to 
> UNSPECIFIED_DOMAIN?  That would seem to break any abstraction 
> between two HPI implementations, and wouldn't be in the 
> spirit of the HPI spec.
> 
> 	-Sean
> 
> --
> __________________________________________________________________
> 
> Sean Dague                                       Mid-Hudson Valley
> sean at dague dot net                            Linux Users Group
> http://dague.net                                 http://mhvlug.org
> 
> There is no silver bullet.  Plus, werewolves make better 
> neighbors than zombies, and they tend to keep the vampire 
> population down.
> __________________________________________________________________
>