-----Original Message-----
From: Unico Hommes Sent: maandag 10 november 2003 12:11
To: [EMAIL PROTECTED]
<snip/>
Sylvain Wallez wrote:
Furthermore, I would like this feature to be defined at the environment level and not only in the flowscript. Methods of the "cocoon" object should just be wrappers around features available also in regular Java code.These methods already exist on the environment objects:
See HttpEnvironment.setStatus(), HttpResponse.setStatus() and HttpResponse.sendError().
Hmm. It seems I was wrong on two accounts here.
First HttpServletResponse.sendError() is specified to send an html error page.
This just means we need to use HttpServletResponse.setStatus(). Only problem is the two argument version - HttpServletResponse.setStatus(int sc, String msg) - seems to be deprecated with no replacement!
The http spec does allow for a customized "reason phrase". So why this was done remains a mystery to me:
-- HTTP/1.1 (RFC 2068) --
6.1.1 Status Code and Reason Phrase
....
The reason phrases listed here are only recommended -- they may be replaced by local equivalents without affecting the protocol.
Status-Code = "100" ; Continue | "101" ; Switching Protocols | "200" ; OK
-- /HTTP/1.1 (RFC 2068) --
What do you think? Should we skip the two argument sendStatus/setStatus methods? Or should we ignore the deprecated warning from the servlet api?
Reading carefully the Javadoc, sendError() and setStatus() are very different:
- sendError() sends an error page and commits the response (nothing more can be sent). The string parameter is the HTML snippet that is displayed in the browser, but not the reason phrase. And this snippet can eventually be overriden if an error page associated to the status code is declared in web.xml
- setStatus() only sets the response status, but does not commit the response, meaning the application can send content afterwards.
And you're right, the servlet API gives us no way to specify the reason phrase (even the deprecated method, in which the string is the same HTML snippet as above). But is it really usefull?
The isRedirected flag still needs to be set on the Redirector upon sendStatus(). We need to decide how. We can do this by specifying null as the url argument to the redirect methods:
package o.a.c.environment;
interface Redirector {
/**
* Redirect to the given URL, null only sets the
* isRedirected flag.
*/
void redirect(boolean sessionMode, String url);
/**
* Unconditionally redirect to the given URL, null only
* sets the isRedirected flag.
*/
void globalRedirect(boolean sessionMode, String url);
}
Or we can add a special nullRedirect() method.
Mmmh... do you mean that we'll have to call both response.sendStatus() and redirector.redirect(null)? This looks hacky...
Moreover, I think we should keep the barriers of Cocoon's response object that forbid access to the underlying response outputstream or also forbid sending redirects from pipeline components. This is why Redirector was created for: allow to those components that have the right to do it (actions, and now flowscript) to send content-less responses.
Up to now these content-less responses were limited to redirects, and I think that sendStatus() would be best located on Redirector.
This would solve all problems at once:
- no possibility to tweak the status from within a pipeline
- have a central object that gathers all methods used to send content-less responses.
WDYT?
Sylvain
-- Sylvain Wallez Anyware Technologies http://www.apache.org/~sylvain http://www.anyware-tech.com { XML, Java, Cocoon, OpenSource }*{ Training, Consulting, Projects } Orixo, the opensource XML business alliance - http://www.orixo.com
