views:

653

answers:

2

Hi.

I've tried to understand this, and searched SO for similar questions, but I still don't have a 100% understanding on how this is supposed to work.

I get this response on a request for an image resource:

Response Headers
    Server  Apache-Coyote/1.1
    Date    Mon, 19 Oct 2009 09:04:04 GMT
    Expires Mon, 19 Oct 2009 09:06:05 GMT
    Cache-Control   public, max-age=120
    Etag    image_a70703fb393a60b6da346c112715a0abd54a3236
    Content-Disposition inline;filename="binary-216-420"
    Content-Type    image/jpg;charset=UTF-8
    Content-Length  4719

The desired behavior is that the client should cache this for 120 seconds, then request it from the server again. Within the 120 seconds, no request is sent to the server.

Then, after 120 seconds, a request is sent and a 304 response is received:

Response Headers
    Server  Apache-Coyote/1.1
    Date    Mon, 19 Oct 2009 09:06:13 GMT

Request Headers
    Host    localhost:8080
    User-Agent  Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.1.3) Gecko/20090824 Firefox/3.5.3
    Accept  image/png,image/*;q=0.8,*/*;q=0.5
    Accept-Language en-us,no;q=0.8,sq;q=0.7,en;q=0.5,sv;q=0.3,nn;q=0.2
    Accept-Encoding gzip,deflate
    Accept-Charset  ISO-8859-1,utf-8;q=0.7,*;q=0.7
    Keep-Alive  300
    Connection  keep-alive
    Referer http://localhost:8080/cms/site/0/en/home
    Cookie  JSESSIONID=768ABBE1A3BFABE3B535900233330650; versionsCssDisplayState=block; iceInfo=iceOn:false,activePortletKey:,icePagePanelX:1722,icePagePanelY:3
    If-None-Match   image_a70703fb393a60b6da346c112715a0abd54a3236

So far, all well. But then, on the next request (whithin 120 seconds) i would have thought that the resource should be cached for 120 new seconds. What i see in the browser (Firefox) on the other hand, is that it from this point on always request the resource and receives the 304-response.

Am I supposed to attach the cache-control headers in the 304-response? From what i can read in the spec, it seems like the cache-control settings should be omitted, and that the cache should cache it for 120 new seconds automatically?

+2  A: 

If I understand correctly then the browser is in fact caching for 120 seconds and your server is responding 304 Not Modified to subsequent If-Modified-Since requests. This "IMS" request occurs when the end-user accesses the same URL. At that time the browser can send an If-Modified-Since request. The browser wants to know if it is displaying stale content. This seems normal.

Upon receiving this request your server should reply 200 OK, 304 Not Modified (or a 4XX, if necessary).

I do not believe you should set your server to send a Cache-Control header with the 304 response for two reasons:
1. You do not want any intermediary caches to cache that 304 response (there is a possibility that they could)
2. The 120 second TTL will not be refreshed by the 304 response. The browser will retain the object for 120 seconds from the 200 OK response. After 120 seconds the browser should send a GET request, not an If-Modified-Since, so your server will respond with the bytes of the file and not just a 304 response.

Note that the browser will not request the file again automatically after 120 seconds unless the end-user specifically requests it via a page load or directly inputting the URL into their address bar (or unless you have a custom application that controls that functionality somehow).

Edited the first paragraph to read a bit better(hopefully)

Structure
Hmmm, maybe a bit unclear from my side, but what i would like to achieve: 1) First request, set cache-control to cache at client for a period of time; max-age. 2) When the max-age is reached, the client should send a new request. 3) If content not changed on server, respond with 304. 4) The browser should cache again for the max-age time. 5) On further requests after max-age, get 304-response and renew the client cache. Now, after the inital 200 and max-age, it "loops" on sending requests to server with 304 as response. Then again, maybe what i would like to achieve isn't a valid approach?
runarM
Here is my take: 1) First request, 200 OK, Cache-Control set via max-age. 2) *If* the client makes a subsequent request after the file is stale, it should make a GET request. Otherwise, an IMS request. 3) If GET request, 200 OK. If IMS, 304 Not Modified or 200 OK. 4) Re-cache file, or retain current copy until max-age is expired and it is requested again, in which case go to #1 above.5) There is no #5. :)If your goal is to have the end-user keep the file as long as possible and only use 304 responses (and you will not update the file) then I suggest a much longer max-age.
Structure
To add to my last comment. I see nothing in the RFC that says your method will not work, or is not valid. However, the wording suggests it is not intended to work that way.The "request" in the RFC suggests a GET request. A 304 is returned in response to an If-Modified-Since request, which is used to validate cache. There is no indication that a Cache-Control header can be included in a 304 response in order to update the headers of a file that is currently in cache. It suggests Cache-Control headers are normally set with the response to a standard GET request.
Structure
If that answered your question then go ahead and mark it as answered. If not then I or someone else will take a stab at clarifying.
Structure
Im still not sure whether its an recommended approach to set the cache-control headers with the 304 response. Ive found examples suggesting this is valid, and the spec isn't very clear on the matter. I'm hoping that someone has some experience in favor of on or the other theory. All my testing so far does, however, indicate that this will work just fine, but then there are a lot of different scenarios to consider.
runarM
There should not be any major downsides. The only considerations I can think of are that intermediary caches may retain the response (but will only do so for 120 seconds). Also, if the end-user makes a subsequent request after 120 seconds, if they wait that long, then you will need to deliver the file bytes again.I still don't see that it is the intended use, definitely appears to be a gray area, but nothing indicates there would be issues. Maybe someone else can lend some input as well.
Structure
+4  A: 

In theory you shouldn't have to send Cache-Control for a 304 -- the recipient should just continue to use the cache directives that it received from the original 200. However, as you've found, in practice if you don't keep sending Cache-Control, browsers will ignore the cache directives that you sent originally, and revert to their own default heuristics.

So in practice, you should include the same Cache-Control with a 304 that you would with a 200. The spec only mandates that you send it for a 304 if it's different than what you sent previously (see 10.3.5 304 Not Modified) -- but it certainly doesn't forbid you from repeating it when it's the same.

And to respond specifically to the wrong-headed points from the other answer (Structure's):

  1. You do want intermediary caches to cache the response (that is, update their cache entry for the resource). They will respond appropriately to requests from clients with a 200 or a 304, depending on whether the client included a conditional header like If-Modified-Since.

  2. The 120-second ttl will be refreshed by the 304 (so the same client shouldn't make another request for the same resource for at least another 120 seconds). And clients, as long as they've still got the content cached, will continue to make conditional requests for the resource, which you can continue to respond to with a 304.

Justin Ludwig