104181 - SH should not store layout state if server sends 'Cache-control: no-cache'

Status: RESOLVED FIXED

(Core :: Networking: HTTP, defect, P1)

Description

[Radha on family leave (not reading bugmail)]

nsIHTTPChannel or nsICachingChannel should have methods that will enable
docshell/session history or any module to inquire about cache specific headers
passed by the server. Something like,

nsIHttpChannel::IsNoCacheResponse().

This is required for bug 104162 and 99692

Comment 1

[Darin Fisher]

Attachment: v1.0 patch -- makes docshell use this method as well

Comment 2

[Darin Fisher]

*** Bug 99692 has been marked as a duplicate of this bug. ***

Comment 3

[Darin Fisher]

ok, this patch makes it so that session history won't remember layout state when
a http server excepts that its response will not be reused by the browser to
satisfy subsequent normal loads (this also includes back/forward, in order to be
consistent with existing web browsers).

Comment 4

[rpotts (gone)]

Comment on attachment 53187 [details] [diff] [review]
v1.0 patch -- makes docshell use this method as well

sr=[e-mail]

I'm assuming that this 'caching' is HTTP specific and shouldn't apply to other nsICachingChannels...

Comment 5

[Darin Fisher]

for the time being, yes.. it is specific to HTTP.. however, in the future it
might make sense to revisit this, and come up with an attribute on
nsICachingChannel that would "be" this for HTTP.

Comment 6

[Darin Fisher]

Attachment: v1.1 revised patch -- forgot one very important check in IsNoCacheResponse

Comment 7

[rpotts (gone)]

Comment on attachment 53216 [details] [diff] [review]
v1.1 revised patch -- forgot one very important check in IsNoCacheResponse

sr=[e-mail]

hey it looks even better now :-)

Comment 8

[Radha on family leave (not reading bugmail)]

r= radha

Darin: I presume the following line in nsHTTPChAnnel.cpp

mResponseHead->Revalidate();

 will take care of pragma: no-cache and cache control:no-cache

Comment 9

[Arun Ranganathan]

So, if Radha's right, we're flushing form fields away with both mechanisms of
using no-cache, both from Cache-Control directive and Pragma directive?

Comment 10

[Jaime Rodriguez, Jr.]

nsbranch+ per Arun, Radha and Darin, and let's talk about it on monday.

Comment 11

[Darin Fisher]

Attachment: v1.2 simple/minimal patch to make us consistent with IE6

Comment 12

[Darin Fisher]

Comment on attachment 53311 [details] [diff] [review]
v1.2 simple/minimal patch to make us consistent with IE6

this patch is incorrect

Comment 13

[Jaime Rodriguez, Jr.]

tentative PDT+, for revised one line path, subject to positive reviews, and
mscott's nod, and wink </*^)

Comment 14

[Darin Fisher]

Attachment: v1.4 simple/minimal patch for trunk

Comment 15

[Darin Fisher]

Attachment: minimal patch for 0.9.4 branch

Comment 16

[Darin Fisher]

these are two minimal patches, which solve this problem, and make mozilla
consistent with the behavior of IE wrt its Back/Forward form history behavior.

the patch for the 0.9.4 branch requires radha's patch to bug 96230.

Comment 17

[rpotts (gone)]

Comment on attachment 53330 [details] [diff] [review]
v1.4 simple/minimal patch for trunk

sr=[e-mail]

Comment 18

[rpotts (gone)]

Comment on attachment 53332 [details] [diff] [review]
minimal patch for 0.9.4 branch

sr=[e-mail]

Comment 19

[Radha on family leave (not reading bugmail)]

r= radha

I tried the latest patch on sites with no-store, no-cache and pragma: no-cache
sites. We are now completely in conformace with IE5.x, but we don't conform with
NS4.x in any way. NS 4.x seems to work in full favor of Pragma: no-cache, which
we don't recognise as much in 6.x.

Comment 20

[Darin Fisher]

yes, 4.x does not know about the HTTP/1.1 Cache-Control header.

Comment 21

[Jaime Rodriguez, Jr.]

rolling the dice, with my bet on Darin and mscott ... let's get this one in tonight.

Comment 22

[Darin Fisher]

patch 53330 checked in on the trunk
patch 53332 checked in on the 0.9.4 branch

resolving FIXED

Comment 23

[Tom Everingham]

verified on 0.9.4 branch - http/forms/cache consistant with IE 5.5

10/15/01 builds  mac osX, Linux rh6, Win NT4