Re: [PATCH 1/3] local-metadata: added a new local source

[Iago Toral <itoral igalia com>]

On Thu, 27 Jan 2011 12:57:37 +0100, Guillaume Emont <gemont igalia com> 
wrote:
> On 27/01/2011 09:05, Iago Toral wrote:
> > Shouldn't we emit an error here? client is using the API with 
> > invalid
> > URI or invalid key...
> >
> >
> > Same here, we should emit an error.
> Not sure where we should draw the line between "no result" and 
> "error".
> But this would be worth defining clearly, ideally in the 
> documentation
> of MetadataSource. For now, Juan seems to say other plugins behave in
> the same way, so I'd be tempted to leave it like this.

I don't think we can draw that line clearly enough, there will always 
be different situations and we cannot know all the reasons a plugin may 
fail for. But I can tell about this one at least. In any case, I think 
the rule should be that we always raise an error unless there is a good 
reason not to.

And as for what Juan says, it is only true for the plugins *he* 
implemented. Check grl-metadata-store for an example of the opposite 
thing ;). Also, if this is wrong, the fact that other plugins do that is 
no reason to justify it, it would be a reason to fix this and the other 
plugins, which is what I think we should do here :)

Iago