[cgiapp] Re: Unexpected variable persistence with CAD / CADS

[George Hartzell]

Mark Stosberg writes:
 > On Tue, 21 Oct 2008 14:02:37 +0100
 > Richard Jones <ra.jones at dpw.clara.co.uk> wrote:
 > 
 > > Richard Jones wrote:
 > > 
 > > > That value of 'id' (2 in this case) persisted from the previous action, 
 > > >  and will continue to live indefinitely across multiple requests until 
 > > > it is either superceded by another id, or until I re-start the CAD server.
 > > 
 > > Behaviour confirmed with a minimal app which just captures id param from 
 > > the url - either using CAP::AutoRunmode - my $id = $self->param('id') - 
 > > or directly using CAP::RunmodeDeclare - runmode foo ($id) - and dumps 
 > > $id to stdout as a warn statement. Once the 'id' param has been used in 
 > > a request, it stays around in $self for subsequent requests, until the 
 > > CAD server is restarted. This seems to me to be incredibly bad and 
 > > dangerous behaviour.
 > > 
 > > It doesn't seem to happen in an identical setup running under mod_perl, 
 > > so I presume it's a 'feature' of the CAD::Server / HTTP::Server. Is this 
 > > a known issue, and one of the caveats alluded to in the CAD::Server 
 > > docs? Never been a problem for anyone else?
 > 
 > I haven't used CAD::Server much. Does it also happen with CGI::Application::Server
 > ( which I think can also work with Dispatch.pm ) ?
 > 

I haven't had time to look into this yet, but hope to get a chance to
poke and prod soon.

In the meantime, I do know that there are patches to C::A::Server
floating around, e.g.

  http://shrimp.alerce.com/cgiapp/

that will make CAD::Server completely redundant.  They're being
reviewed by the maintainer, haven't heard any feedback.

g.