Back button broken

[GoogleCodeExporter]

What steps will reproduce the problem?
1. Go to a page with some interactive elements
2. Navigate away from that page
3. And then hit the back button to go back to the original page 

What is the expected output? What do you see instead?
The interactive elements disappear.

This could be tricky to fix - a lot more investigation is needed.

Original issue reported on code.google.com by johannes...@gmail.com on 24 Nov 2007 at 11:10

[GoogleCodeExporter]

By the way, I found this using Camino - but it should also hold for firefox.

Original comment by johannes...@gmail.com on 24 Nov 2007 at 11:10

[GoogleCodeExporter]

I believe that this is impossible to fix - the communication link between the 
page
(javascript code) and the python server gets broken.

Original comment by andre.ro...@gmail.com on 25 Nov 2007 at 3:36

[GoogleCodeExporter]

Set as invalid - I believe this is totally beyond the control of Crunchy.

Original comment by andre.ro...@gmail.com on 30 Dec 2007 at 10:56

• Changed state: Invalid

[GoogleCodeExporter]

Perhaps I spoke too soon.  See:
http://www.onjava.com/pub/a/onjava/2005/10/26/ajax-handling-bookmarks-and-back-b
utton.html

Original comment by andre.ro...@gmail.com on 30 Dec 2007 at 10:59

• Changed state: Accepted

[GoogleCodeExporter]

This is still an issue, but I think I know why: the page is being cached by the
browser and its not reloaded when you hit the back button. Should be relatively 
easy
to fix by sending some specific header to the browser when it loads pages, 
telling it
not too cache.

Original comment by johannes...@gmail.com on 1 Jun 2008 at 8:22

[GoogleCodeExporter]

I've made some search on the web, and find it is possible to make Firefox (and 
maybe
IE) reload the page when you go back with 'Cache-Control' http header. The 
header is:

Cache-Control: no-cache, must-revalidate, no-store

I've tested it on various browser. It works well on gecko-based browsers 
(Firefox,
Epiphany), but not in Opera or Konqueror (they seems to be more respectful to 
the
HTTP specification, which say that history is not a cache).
My IE running on Wine doesn't want to log in Crunchy, so if anyone can try with 
IE?

I've committed the modification on my branch (r806).

Original comment by florian....@gmail.com on 1 Jul 2008 at 1:46

[GoogleCodeExporter]

Florian:

Good work.

Your solution does indeed conform to the idea expressed in Comment 5 by 
Johannes. It
is probably the best we can achieve.

What I had originally been thinking, when Johannes raised the issue, is that we
should find a way to both make the lost interpreter reappear (which your 
solution
does) AND keep the history of commands entered (i.e. not lose the calculations),
rather than reloading the page (which, according to what you wrote, is in 
agreement
with the HTTP specification).  I will leave the issue open until I incorporate 
it in
the trunk, but otherwise consider it fixed.

As for IE not working... IE was not working for the past two years with 
Crunchy.  
Perhaps the newest version of IE works... but we won't bother with it until we 
move
to using jquery (which should take care of all browser incompatibility issues 
for us).

Original comment by andre.ro...@gmail.com on 1 Jul 2008 at 2:17

[GoogleCodeExporter]

With Florian's latest contribution incorporated in the trunk, I'm changing the 
status
to "fixed".

Original comment by andre.ro...@gmail.com on 8 Jul 2008 at 11:17

• Changed state: Fixed