Ian Bicking wrote:
>On Fri, 2003-01-10 at 17:38, Stuart Donaldson wrote:
>>I am encountering numerous problems trying to use the psp-handler with
>>mod_webkit, and some of the more recent code in Webware.
>>The PSP handler in mod_webkit sets WK_ABSOLUTE=1 to indicate that the
>>request is not relative to one of the webware directories.
>>HTTPRequest urlPath() checks this, and returns the fsPath() or file
>>system path under this circumstance. That appears to be the only case
>>where the urlPath() returned is a file system path rather than web
>>server url component.
>>The serverSideInfoForRequestNewAlgorithm does not check for the absolute
>>condition, and tries to process the filesystem path as a url, passing it
>>off to findContext() et al... The old algorithm did check for this by
>>Is there ever a circumstance where urlPath() should really be returning
>>a filesystem path rather than a URL component? It seems like code which
>>wants to check for absolute paths from the psp-handler should probably
>>look specifically at wether or not the request is an absolute request,
>>and then use fsPath() instead. One of the with urlPath returning a file
>>system path is that it is often combined with other components of a URL
>>to generate new URL's.
>>It appears that this general approach of using the psp-handler is really
>>broken in the New functions, and I just wanted to check to see if the
>>psp-handler approach is something we want to continue to support, or if
>>it is being phased out. I'm kind of surprised it hasn't come up much
>It isn't being phased out, but I must admit I forgot about it when
>implementing those functions, and I feel a little unclear about how it
>should work -- it's something I haven't used. Well... we get four items
>from the URL parsing: context path, context name, servlet path, and
>extra URL path.
I am currently digging into these problems some more, but one difficulty
is the multiple code paths based on the variety of configurations
It appears that some of the problems are related to parsing the
environment in HTTPRequest. The adapterName with a psp-handler request
is always the actual psp file. This can cause problems for anything
that goes through setURLPath such as callMethodOfServlet and includeURL
because they kind of assume the adapterName is the URL prefix which
redirected into Webware.
This is also related to some of the problems reported in SF, and on the
mailing list regarding includeURL and reltive vs absolute paths.
>>Perhaps the problem is that I'm also working with code in
>>NewThreadedAppServer which may not be getting the broad testing we would
>No, this is a problem with the new path parsing code (I assume you are
>using ExtraPathInfo). It should be shared between both code bases, and
>the bug should be in Application, not any of the AppServer classes.
Yes, I am using the ExtraURL. Should the ExtraURLPathInfo be on by
default? Or possibly even removed eventually? Why would you not want
the extra information?