748108 - Incorrectly caching redirects that lack an Expires header.

Status: RESOLVED WONTFIX

(Core :: Networking: Cache, defect)

Description

[Jason Duell]

Attachment: nc script

Noticed this while debugging.  We're storing 3xx requests with a cache entry expiration time of 0 (unix epoch).  Then when we hit the same URI, we load the entry, notice it's old, and wind up hitting the server anyway.  This wastes time (useless I/O) and space.

I suspect there's both an HTTP and cache solution to this.  HTTP shouldn't write a cache entry with expires=0 time on it (we should be writing "no expiration" IIUC).   And the cache should probably be changed to automatically doom any entry that has an expired cache time of 0.  

Steps to reproduce.  Run
  
  nc -k -l 8080 <vanilla_redirect_ppl

hit "localhost:8080/foo" once.  Open about:cache and you'll see entry with expires=0.   Hit it again and you'll see that the nc server gets hit again.

If you change the nc file to contain
 
   Expires: Thu, 19 Nov 2014 08:52:00 GMT

Then we store the entry with that date, and we don't wind up asking the server for the redirect when we hit it again.

Comment 1

[Patrick McManus [:mcmanus]]

we store just about everything for offline, etc..