Cache feature doesn't produce expected result (bug?)

[Tom.PesterDELETETHISSS]

Hi,

I think this question requires an in depth understanding of how a browser
cache works. I hope I can reach an expert here.

I may have found a quirk in the asp.net documentation or I don't understand
what the SetAllowResponseInBrowserHistory does.
While researching caching I tried the code sample at the following page :

http://msdn2.microsoft.com/library/97wcd0a4(en-us,vs.80).aspx

I find this code absurd since it tries to show that you can view a page that
has been _posted_ with the history feature of the browser.

Can you try this code and report your findings please? I tried it on a clean
VM install and on my dev pc: the browser says "page expired".
The docs say you would see the old page in the browser.

I did some tests and this is what the following code does with the response
headers :

Response.Cache.SetCacheability(HttpCacheability.NoCache)
Response.Cache.SetAllowResponseInBrowserHistory(False) ' False
is the default
'Cache-Control: no-cache
'Expires: -1
'Pragma: no - Cache

Response.Cache.SetCacheability(HttpCacheability.NoCache)
Response.Cache.SetAllowResponseInBrowserHistory(True)
'Cache-Control: no-cache
'Pragma: no - Cache


The docs are right about the Expires header. When SetAllowResponseInBrowserHistory
= True the Expires header disappears.

Does anybody have a clue why SetAllowResponseInBrowserHistory was called
into existence and where it might be useful?

Thanks,
Tom Pester

 

[Juan T. Llibre]

re:

Does anybody have a clue why SetAllowResponseInBrowserHistory was called into existence
and where it might be useful?

SetAllowResponseInBrowserHistory(false) is the default.

You only need to set HttpCachePolicy.SetAllowResponseInBrowserHistory
when you need to set it to true to override the NoCache setting

 

[Tom.PesterDELETETHISSS]

Hi Juan,

Thanks for your response.

You only need to set HttpCachePolicy.SetAllowResponseInBrowserHistory
when you need to set it to true to override the NoCache setting

Setting SetAllowResponseInBrowserHistory to True doesn't touch/override the
"no-cache" value. It only removes the expires=-1 header.

This code shows this :
Response.Cache.SetCacheability(HttpCacheability.NoCache)
Response.Cache.SetAllowResponseInBrowserHistory(True)

The servers send the following cache related headers :
'Cache-Control: no-cache
'Pragma: no - Cache

I don't know if you are experienced in http headers and their behavior but
could you please test the code that's on this page and report your result? :

http://msdn2.microsoft.com/library/97wcd0a4(en-us,vs.80).aspx

If you run the sample and go back with your browser I think you will find
that the text on the page is wrong.

Thanks in advance, Tom

 

[Juan T. Llibre]

re:

Setting SetAllowResponseInBrowserHistory to True doesn't touch/override the "no-cache"
value. It only removes the expires=-1 header.

Maybe we have a semantics problem here.

Removing the expires= -1 header *equals* overriding the "NoCache" value.

When HttpCacheability is set to NoCache or ServerAndNoCache
the Expires HTTP header is set to -1 by default.

NoCache and ServerAndNoCache instruct the client to
not cache responses in the History folder by setting that header.

This means that each time you use the back/forward buttons,
the client requests a new version of the response.

When SetAllowResponseInBrowserHistory is set to True,
the Expires HTTP header is removed.

If you comment out this line :
HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(True)
and alternate between clicking the Submit button and the Back button,
you'll see that the "This page has expired" message is not displayed.

That means that the client *has* requested a new version of the page,
without having to resubmit the page.

If you set
HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(False)
you'll see that immediately the "This page has expired" message *is* displayed,
and the client needs to resubmit the page.

Now, if you set
HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(True)
and alternate between the Submit button and the Back button,
you'll see that the "This page has expired" message is not displayed,
and the page is displayed without needing to resubmit the page.

The documentation is wrong in requesting that you
"Click the Submit button a few times".

That throws a wrench into the works.
You should only hit it once to see the correct behavior.

The documentation is also wrong when it states that
SetAllowResponseInBrowserHistory allows client-side caching.

In effect all it does is remove the need to resubmit the page.

I hope this makes this issue clearer.

 

[Tom.PesterDELETETHISSS]

Hi Juan,

I tried exactly the steps that you advised but this is where I can't reproduce

If you set HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(False)

you'll see that immediately the "This page has expired" message *is* displayed,
and the client needs to resubmit the >page.

I dont see the "This page has expired" message and if I comment out SetAllowResponseInBrowserHistory
I get the same result as False is the default. No "This page has expired"
message, the browser request a new page as I can see by the time.

Could it be that you saw this when testing without restarting IE? I even
made a video of my clean virtual machine (just SP2 installed) to show you
my result.

Please try this:
http://users.pandora.be/TomPester/ASP/r.rar

Do you realy get a "This page has expired" when you do what's on the video?
I don't.

Setting SetAllowResponseInBrowserHistory to true or false... I never see
a difference.

Cheers,
Tom Pester

 

[Juan T. Llibre]

re:

Could it be that you saw this when testing without restarting IE?

I didn't restart IE, thinking that editing the source file would
automatically force recompilation ( and return a new page anyway ).

Let me take a look at it while closing IE, and I'll post back.

 

[Tom.PesterDELETETHISSS]

Thx for sticking with this Juan. I dont have an urgent problem with this
feature, I just whant to know everyting there is about caching cause I neglected
it for too long.
This SetAllowResponseInBrowserHistory is one of the last mysteries for me.

Thanks,
Tom Pester

PS I am using ASP.NET V2 Beta 2 so the page recompiles every time I make
a change.

 

[Juan T. Llibre]

OK...

First test :
HttpContext.Current.Response.Cache.SetCacheability(HttpCacheability.NoCache)
and
HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(True)

Results :
Browser doesn't produce "Page has expired" message
unless the submit button is clicked more than once.

Browser does produce "Page has expired" message
if the submit button is clicked more than once.

Single-click behavior returns different time stamps for each page.

Browser closed.

Second test :
HttpContext.Current.Response.Cache.SetCacheability(HttpCacheability.NoCache)
was commented out
HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(True)
is left in code

Results :
Browser doesn't produce "Page has expired" message
unless the submit button is clicked more than once.

Browser does produce "Page has expired" message
if the submit button is clicked more than once.

Single-click behavior returns different time stamps for each page.

( same results as the first test )

Browser closed.

Third test :
HttpContext.Current.Response.Cache.SetCacheability(HttpCacheability.NoCache)
and
HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(False)

Results :
Browser doesn't produce "Page has expired" message
unless the submit button is clicked more than once.

Browser does produce "Page has expired" message
if the submit button is clicked more than once.

Single-click behavior returns different time stamps for each page.

( same results as the first and second tests )

Browser closed.

Fourth test :
HttpContext.Current.Response.Cache.SetCacheability(HttpCacheability.NoCache)
was commented out
HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(False)
was left in code.

Results :
Browser doesn't produce "Page has expired" message at any time.

Browser allows returning to previous versions of page ( as determined
by the date stamp ) no matter how many times the submit button is
pressed successively.

Whew !

Analysis :
The only time the code works as expected is when
HttpContext.Current.Response.Cache.SetCacheability(HttpCacheability.NoCache)
is commented out and
HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(False)
is left in the code.

My conclusions :

1. setting NoCache invalidates SetAllowResponseInBrowserHistory
regardless of whether SetAllowResponseInBrowserHistory is set to True or False.

2. If NoCache isn't set but SetAllowResponseInBrowserHistory is set to True
the response is the opposite of what it should be. I would expect SetAllowREsponse
InBrowserHistory to, precisley, allow the browser to use its history cache.

3. if SetAllowResponseInBrowserHistory is set to False *and*
SetCacheability(HttpCacheability.NoCache) is *not* set
then everything works, sort of, since the result is exactly the opposite
of what I would have expected.

It seems that the functionality of SetAllowResponseInBrowserHistory
works if set to False instead of being set to True, but only if NoCache
isn't set previously.

This is very strange, but I've seen stranger.

I will escalate this up the line to PSS, and try to get a reply.
When I do, I'll post it back.

Before doing that, I'll wait for any last comments you might have.
Thanks for pointing this out.

 

[Tom.PesterDELETETHISSS]

Hi Juan,

I want to point out that History mechanisms and caches are different in case
you didn't know.
You can read about it in section 13.13 at http://www.faqs.org/rfcs/rfc2616.html
Although I think IE doesn't follow this section ( a cache-control:no-cache
does mean a request to the server if you hit back), there could be some semantic
differences we have to take into account when talking about the history feature
and the browser cache.

1. setting NoCache invalidates SetAllowResponseInBrowserHistory regardless

of whether SetAllowResponseInBrowserHistory is set to True or False.

I watched the headers going back and forth and SetAllowResponseInBrowserHistory
= True only takes away the expires=-1 header if SetCacheability is set to
NoCache.
But this has no effect on the browsers behavior. All your tests returned
the same result as well as mine. So I think the question is what does expires=-1
try to do and why would you want to get rid of it?

2. If NoCache isn't set but SetAllowResponseInBrowserHistory is set to True
the response is the opposite of what it should be. I would expect SetAllowREsponse
InBrowserHistory to, precisley, allow the browser to use its history cache.

if I Make SetAllowResponseInBrowserHistory True than I _do_ get the page
from the cache but this is actually the default behavior of asp.net :
the cache-control header is set to private and the SetAllowResponseInBrowserHistory
could as well be commented out. It doesn't add or change any headers on its
own.

You get a fresh page?

3. if SetAllowResponseInBrowserHistory is set to False *and*
SetCacheability(HttpCacheability.NoCache) is *not* set
then everything works, sort of, since the result is exactly the opposite
of what I would have expected.

The cached page is displayed right? But this contradicts SetAllowResponseInBrowserHistory=False
or its confusing at least.

It seems that the functionality of SetAllowResponseInBrowserHistory works

if set to False instead of being set to True, but only if NoCache isn't set
previously.

So you have found a page where setting the SetAllowResponseInBrowserHistory
to True or False did influence the outcome??
My test say that SetAllowResponseInBrowserHistory doesn't do anything that's
noticeable for the end user
It does however remove the Expires=-1 header when you set SetCacheability
to HttpCacheability.NoCache but like I said this doesn't effect a thing in
my findings.

I will escalate this up the line to PSS, and try to get a reply. When I

do, I'll post it back.

Seems I found the right person Asking the right question is important.
This is what I would ask:

Can you give a code sample where SetAllowResponseInBrowserHistory does have
an effect on the end user experience?

Notice that I don't ask what SetAllowResponseInBrowserHistory is supposed
to do because after so many tests I don't make any assumption on what it
is trying to accomplish.
I hope that with a code sample we can see the effect ourselves and draw conclusions
from it.

Cheers,
Tom Pester

 

[Tom.PesterDELETETHISSS]

Hi Juan,

To come back to your fourth test that differed from the previous ones:
"Fourth test : HttpContext.Current.Response.Cache.SetCacheability(HttpCacheability.NoCache)
was commented out HttpContext.Current.Response.Cache.SetAllowResponseInBrowserHistory(False)
was left in code."

If you dont use SetCacheability on a page than ASP switches back to its default
cache-control wich is private, ie the browser may cache the page. So you
can press the submit button many times and still go back in history.
Setting SetAllowResponseInBrowserHistory to False of True has no effect at
all on the headers so it could as well be commented out.

Can you report this issue a level higher please?

Let me know if you have any more questions..

Cheers,
Tom Pester

 

[Juan T. Llibre]

I just verified by commenting out
SetAllowResponseInBrowserHistory

btw, what are you using to see the headers ?

re:

Can you report this issue a level higher please?

Will do so...and post back as soon as I get an answer.

 

[Tom.PesterDELETETHISSS]

There are a few excellent freeware/shareware tools that you can use to debug
:

- http://www.blunck.info/iehttpheaders.html
- http://www.fiddlertool.com/fiddler
- http://www.proxomitron.info

Some more links :

- http://www.ieinspector.com/httpanalyzer
- http://www.iewatch.com
- http://www.httpsniffer.com/screen.htm
- http://www.httpwatch.com/viewer.htm

Firefox has a good extension too.

Cheers,
Tom Pester

 

[Juan T. Llibre]

Hi, Tom.

OK, I've been using

http://www.blunck.info/iehttpheaders.html

Just wondered which one you're using ( you didn't tell me... )

;-)

 

[Tom.PesterDELETETHISSS]

Proxomitron cause I use it for other stuff too


Cheers,
Tom Pester

 

[Tom.PesterDELETETHISSS]

Hi Juan,

Any news from the folks at MS?

Cheers,
Tom Pester