Possible removal of ctx from inevow.IResource (was Re: [Twisted-web] access control feature, ctx removal from locateChild and renderHTTP, what's going on?)

[Jp Calderone]

On Sat, 03 Sep 2005 10:03:16 +0100, Matt Goodall <matt at pollenation.net> wrote:
>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>Tommi Virtanen wrote:
>
>[snip]
>
>> <exarkun> Tv: it'll break when ctx is removed as an argument to
>> locateChild et al
>> ...
>> <Tv> exarkun: err, wouldn't a _lot_ break with that change?
>> <exarkun> Tv: yes!  but this, too.
>> <Tv> well, I can make it give a wrapper resource from locateChild
>> <Tv> that does the access checking at render time..
>> <exarkun> ctx will be removed from renderHTTP at about the same time
>> <Tv> so huh
>> <Tv> 1) what's the reasoning behind that
>> <Tv> 2) what, then?
>> <exarkun> those questions are big and I have to catch a bus
>
>Jp, I assume you caught you bus now ;-) so please could you post your
>ideas about removing the context arg from IResource.
>
>That is a *major* change that would affect every single Nevow
>application so it needs proper discussion, including how to handle
>backwards compatability.
>

This is a very long term plan.  There will be no changes of this sort in the immediate future.  When the changes are made, there will be a very clear migration path.  Since the ways in which the context can be used are legion, I am not sure if there will be a completely backwards compatible API layer, but this will be the goal for as long as possible.

When the time comes to make this change, there will be an opportunity for more discussion (it won't just be mentioned in passing on IRC and then done willy nilly).  I'd welcome any ideas or proposals anyone has now, but I doubt I'll have time to participate in any discussion that happens in the foreseeable future.

Jp