I don't think the patch came through because I'm guessing the mailing
list is stripping all attachments :/
Also, I want to be fairly careful about the decision to just add the RSD
support to the FlavorServlet. I still prefer the RSD servlet idea.
My reason is derived mainly from the fact that I believe that there are
various places in Roller where we have tried to squeeze too much
functionality into a small number of classes/objects. There are also
places where we have tried to create generic all purpose code rather
than taking the additional time and effort to separate out functionality
that should be a bit decoupled.
The two examples I think are most relevant are the RollerRequest object,
which in my opinion is a good idea, except that the way it's
impelemented is a bit too much "all in one." I think the RollerRequest
can be broken down into smaller more specialized subclasses which would
be more effective. Another example is the way we setup the Velocity
Context object, where we basically try and stuff all things that any
page would want (page, search, flavor, etc) rather than having a
separate context setup process per function.
I think the RSDServlet is a good place to start to break this habit by
being very lightweight and only hitting the db for things it really
needs, and also by creating it's own velocity context populated with
only the items it needs.
As usual, others opinions are welcomed.
-- Allen
On Wed, 2005-12-14 at 17:50, Sean Gilligan wrote:
> I went ahead and attached an "svn diff" (I'm not sure that's the right
> way to submit a patch, let me know if it isn't) to ROL-930.
>
> This patch fixes ROL-899, ROL-930 and also uses
> #showContentType("text/html; charset=utf-8") in place of a direct <meta>
> tag for consistency between templates.
>
> Further, some files had CRLF instead of LF, so I changed/fixed that, too.
>
> If we later decide to write an RSDServlet, we can change the
> #showRSDLink() macro and the theme templates won't have to change.
>
> Regards,
>
> Sean
>
>
