Bertrand Delacretaz wrote:
Le 9 avr. 05, � 19:21, Sylvain Wallez a �crit :
...Cocoon is sometimes too much powerful and people use too much "cocoon:" dynamic URLs when other systems would require to dump something on disk, and this can lead to overly complex and and very unefficient architectures.
So, although we may accept a new "context" attribute (I'm currently -0.9 for this), I'm -1 for accepting dynamically generated sitemaps...
Hmm...sounds a bit like the arguments against XWindows when it was created: at that time most people couldn't afford the system and network resources used to run it. But the "processors and networks get faster all the time anyway" attitude made XWindows a very long-lasting and powerful system.
So, although dynamically generated sitemaps might be risky performance-wise today, I don't think we should block them. People might find brilliant uses for them - we just have to warn our users about potential bottlenecks.
Well, at work we currently have to maintain and evolve a project (written elsewhere) where people have written a gazillion chained pipelines with "cocoon:" everywhere where a simple Java class would have done the job. A nightmare to understand and maintain.
And seeing <map:mount src="cocoon:/blah"/> seems to open dangerous doors to me...
That being said, there's another issue about the proposed "context" attribute: who's responsibility is it to specify on what context will a sitemap operate? Is it the <map:mount>'s responsibility or the sitemap's own responsibility?
In other words, should it be:
<map:mount src="foo/bar/" context="baz"/> and <map:sitemap>
or <map:mount src="foo/bar/"/> and <map:sitemap context="baz"/>
My feeling favors the second form, as allowing to specify the context externally means that understanding what a sitemap does depends on the mount statement.
Well, that's right that a sitemap already depends on the mount statement for component definitions inherited from the parent. But now knowing what "blah.xml" means in <map:generate src="blah.xml"/> seems really weird to me...
Maybe I'm just too conservative about this, but this doesn't smell good...
Sylvain
-- Sylvain Wallez Anyware Technologies http://apache.org/~sylvain http://anyware-tech.com Apache Software Foundation Member Research & Technology Director
