Re: Response Fixes

2014-07-02 Thread Łukasz Rekucki
ectations, auto-fix is auto-break. > > We recently removed two of the four unconditional response "fixes". > > The remaining ones are: > > - `fix_location_header`: at least for some people, it's an issue. That's > why we're having a discussion. > > - `condition

Re: Response Fixes

2014-07-02 Thread Łukasz Rekucki
On 2 July 2014 15:42, Florian Apolloner wrote: > On Wednesday, July 2, 2014 3:36:22 PM UTC+2, Łukasz Rekucki wrote: >> >> It doesn't just alter it, but makes it conform to HTTP standard. While >> most browsers will accept relative urls, I don't think Django should >>

Re: Response Fixes

2014-07-02 Thread Aymeric Augustin
2014-07-02 15:36 GMT+02:00 Łukasz Rekucki <lreku...@gmail.com>: It doesn't just alter it, but makes it conform to HTTP standard. As usual, given a different set of expectations, auto-fix is auto-break. We recently removed two of the four unconditional response "fixes". Th

Re: Response Fixes

2014-07-02 Thread Florian Apolloner
On Wednesday, July 2, 2014 3:36:22 PM UTC+2, Łukasz Rekucki wrote: > > It doesn't just alter it, but makes it conform to HTTP standard. While > most browsers will accept relative urls, I don't think Django should > sacrafice backwards compatibility for an arcane CGI feature. Internal

Re: Response Fixes

2014-07-02 Thread Łukasz Rekucki
On Jul 2, 2014 2:09 PM, "Aymeric Augustin" < aymeric.augus...@polytechnique.org> wrote: > > I find it wrong to alter the response created by the developer unconditionally and not provide any escape hatch. It doesn't just alter it, but makes it conform to HTTP standard. While most browsers will

Re: Response Fixes

2014-07-02 Thread Aymeric Augustin
I find it wrong to alter the response created by the developer unconditionally and not provide any escape hatch. Therefore option 1 is my favorite. I'm proposing to documenting the backwards incompatibility in the release notes. It will only affect project that do not use CommonMiddleware. There

Response Fixes

2014-07-01 Thread peter
This is in reference to this ticket: https://code.djangoproject.com/ticket/17092 There is a patch there to fix the specific use case of needing to disable django's fix_location_header for certain responses in a CGI compliant environment. Because of the way that response_fixes work you can't