Carsten Ziegeler wrote: >Ok, this seems to be a very difficult task then. > >Let me state the following: >a) There shouldn't be two places where one can define the sitemap >components, > so it's rather the xconf or the sitemap. But having both is absolutly > confusing. > I tend to agree with this. I consider cocoon.xconf to be the location for "system" components, i.e. components that are available for _all_ components, be them related to a sitemap or not. But we may consider some sitemap components as system-level (see below).
> >b) Having said a), if we decide for the xconf, it will *not* be possible > to define custom components in sub-sitemaps! > Definitely -1 for that. We *need* components defined locally for a sitemap. The treeprocessor even offers more features for that than the compiled engine, since the <map:component> is a regular component manager configuration where you can add *any* component definition (e.g. local jdbc datasources). However, I agree that component definition can be considered as not being the concern of the sitemap administrator (he/she manages the URI space, not the application configuration). So what about allowing <map:components> to be written <map:components src="sitemap.xconf"/> ? This way, each [sub-]sitemap (or cocoon block) comes with a sitemap.xmap and an optional sitemap.xconf defininig its local components, if any. The current behaviour should also be kept, for compatibility and for people which aren't confused by having components in the sitemap. Now the question is do we define system-wide sitemap components in cocoon.xconf ? If yes, they should be IMO restricted to the very minimum of what we're likely to find in _every_ sitemap, i.e. wildcard and regexp matchers, file and xsp generators, xsl transformer, xml and html serializers. Note also that some system-level components may make use of some sitemap components. Namely, the xml serializer used in some source implementations, IIRC. Thoughts ? > >c) If we opt for defining the components in the sitemap (as it is now) > we help the sitemap editor in writing the pipelines as he can simply > see which components are available. > >So, SoC or not, is the above really what we want? I think, no! > >Hmm, currently I'm thinking of voting -1 for defining the components >in the xconf. This would create a deadlock. Very interesting >and funny thing... > Does sitemap-local xconf remove the lock ? Sylvain -- Sylvain Wallez Anyware Technologies Apache Cocoon http://www.anyware-tech.com mailto:[EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, email: [EMAIL PROTECTED]
