188610 - Recently created content with no expiration in the response header expires almost immediately

Status: RESOLVED WORKSFORME

(Core :: Networking: HTTP, defect)

Description

[David M. Pochron]

User-Agent:       Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.0.3705)
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.3a) Gecko/20030110

Pages such as yerf.com that have new content constantly being added (in this 
example a listing of recent images) and this new content has no expiration time 
makes this content expire from the cache too soon because the freshness value 
is based on the last-modified time of the conent.  (In this case, this means an 
almost immediate expire time.)

This means content that hasn't changed must be transferred from the web site 
every time the page is revisited, causing a major performance issue for the 
majority of the world which doesn't have high-speed internet access.


Reproducible: Always

Steps to Reproduce:
1. Set cache updating to "once per session".  (always may work, too.)
2. Go to yerf.com/recent
3. Allow all of the image thumbnails to load.
4. Go to about:cache and note the expiration times of the most recently added 
images.
5. Wait until some of the expiration times you noted pass.
6. Close and reopen the browser. (Make sure you don't have your cache updating 
set to "never".)
6. Revisit the recent images page again.
Actual Results:  
If you have a slow connection you'll notice the images that have expired must 
all be transferred again.


Expected Results:  
Cached content with no expiration time shouldn't expire immediately and should 
be fetched from the cache instead of the web server.

This bug is also in Mozilla 1.2.1.  (In fact, it's in every release since the 
new caching method was applied.  I suspect it didn't appear earlier because the 
expiration routines weren't working properly with the old cache, so this 
problem was simply not manifesting itself.)

This bug is related to 80731, however that bug doesn't bring up the fact that 
pages expire immediately because they don't have an expire time - it only talk 
about how users sohuld be able to control the lifetime of content in the cache.

While taking care of bug 80731 would make this problem go away, I don't think 
the HTTP networking code makes a good assumption about how to calculate a 
freshness values, despite what RFC2616 says.  I think they assumed that content 
with a recent modification time is content that is being generated dynamically, 
which is not the case here. (Nor with most sites that don't use expiration 
times but have recently updated their content - and this is a large chunk of 
the Internet!)

For what it's worth, Internet Explorer doesn't expire content that lacks an 
expiration time.

Comment 1

[David M. Pochron]

Well..  Now it's working.  Perhaps the server wasn't responding to if-modified
conditionals correctly.  Also, I got validation and expiration mixed up:
Expiration & max-age only specify whether the browser will *attempt* to do a
revalidation rather than retransferring the data unconditionally.