On Dec 13, 2005, at 1:22 PM, David Binger wrote: > On Dec 13, 2005, at 5:20 AM, mario ruggier wrote: > >> What would the arguments be against a init_response=True param to >> pub.redirect() > > I try to resist adding keyword arguments where the value of the keyword > is, from the caller's point of view, a constant. The "permanent" > keyword > that is already in redirect, is an example of what I try to resist: > the caller > would always know in advance if this is a permanent or temporary > redirect. I'd rather just have a separate redirect_permanent() method > because I think this makes it easier to understand the behavior of > the method. Adding an init_response keyword makes this situation > more extreme. I agree with the resistance to parameter proliferation. But function/method proliferation is a similar problem, or worse if you start getting into combinations of possibilities... What would win it for me is how easy can I remember what's available, and which reads clearer. In the case of permanent, as it seems to be used in very clear context, it seems good to have a redirect_permanent(). > I think your use-case is common enough to justify some support from the > Publisher. I'm trying to figure out the best way to do that. > Is it correct that it is always set-cookie header(s) that one > would want to preserve? > Should we just delete the init_response() call? My feeling is that all other http headers will always be set during normal response processing, i.e. not during a redirect.... on the other hand, it does seem reasonable to do the systematic defensive call of init_response() on redirects, as you just do not want to worry about potential problems from having things set on the response prior to the redirect, that could then cause conflicts. Is there such a potential problem? Anyway, what was the reasoning for re-initialising the response on redirects in qp (differs from quixote) ? m.