- From: Mark Nottingham <mnot@mnot.net>
- Date: Thu, 4 Mar 2010 15:39:59 +1100
- To: HTTP Working Group <ietf-http-wg@w3.org>
- Cc: Alex Rousskov <rousskov@measurement-factory.com>, Roy Fielding <fielding@gbiv.com>
We haven't heard back from Alex, and the other issue I mentioned didn't seem to get enough support to move on. So, I suggest we do the conservative thing: Current text: > age_value - Age header field-value received with the response > date_value - Date header field-value received with the response > request_time - local time when the cache made the request > resulting in the stored response > response_time - local time when the cache received the response > now - current local time > > apparent_age = max(0, response_time - date_value); > corrected_received_age = max(apparent_age, age_value); > response_delay = response_time - request_time; > corrected_initial_age = corrected_received_age + response_delay; > resident_time = now - response_time; > current_age = corrected_initial_age + resident_time; Replacement text: > age_value - Age header field-value received with the response; > 0 if not available. > date_value - Date header field-value received with the response; > see [ref] for requirements regarding responses > without a date_value. > request_time - local time when the cache made the request > resulting in the stored response > response_time - local time when the cache received the response > now - current local time > > apparent_age = max(0, response_time - date_value); > response_delay = response_time - request_time; > corrected_initial_age = max(apparent_age, age_value + response_delay) > resident_time = now - response_time; > current_age = corrected_initial_age + resident_time; Comments? On 14/10/2009, at 8:31 PM, Mark Nottingham wrote: > Hi Alex, > > We put this on the HTTPbis issues list a while ago <http://trac.tools.ietf.org/wg/httpbis/trac/ticket/29>, and I've discussed it with a few folks F2F, but AFAICT it hasn't been discussed on-list. > > In a nutshell, I think you're correct that there's a problem here, but your proposal: > >> creation_time = min(date_value, request_time - age_value); >> current_age = now - creation_time; > > has a few (small-ish) issues. > > 1) The corrected_received_age's subtraction of the date_value from now has the (intended, I assume) effect of accounting for upstream HTTP/1.0 caches that don't append an Age header. Your proposal doesn't do this. > > This is already being diccussed on-list (see recent thread "cache freshness / age calcs"), and may go away anyway. Your input there would be appreciated. > > 2) The behaviour when date_value isn't present isn't specified; we could address this in prose, but it would be awkward. > > This could probably be worked around by either specifying a slightly more complex formula, or specifying that when the Date header isn't present, a completely separate (and presumably much simpler) formula is to be used. > > Cheers, > > > > On 31/08/2002, at 2:59 AM, Alex Rousskov wrote: > >> >> Hi there, >> >> We are testing a couple of RFC 2616 MUSTs related to >> current_age calculation. Many proxies violate a subset of test cases >> that includes an artificial proxy-to-server delay. Looking at the >> results, I think that the proxies are doing the "right thing" and the >> RFC has a problem. >> >> I will start with a specific example when current_age formula >> from the RFC yields a way-too-conservative and unnatural result (100% >> error). I will then describe the problem and suggest a fix. >> >> I understand that a way-too-conservative age does not lead to >> stale documents being returned. However, if we want proxies to be >> compliant, we may want to fix/mention the problem in the errata or >> elsewhere. Otherwise, the more problems like that are left unaddressed >> (ignored), the more difficult it would be to convince implementors to >> pay attention to the RFC. >> >> Perhaps I got it all wrong, please check! >> >> >> A simple example >> ---------------- >> >> Here is a real and simple example that detected the problem with the >> original current_age formula from "13.2.3 Age Calculations". The >> absolute values of timestamps below ("0" and "7") have no >> significance. >> >> time event >> ---- ------------------------------------------------------------ >> 0.0 client request generated >> 0.0 client request reached the proxy, it is a MISS >> 0.0 proxy request to origin server is generated >> 0.0 proxy request reached the origin server >> 0.0 server response generated with Date correctly set to 0, no Age header >> -- a network delay of 7 seconds -- >> 7.0 server response reached the proxy >> 7.0 proxy cached the response >> 7.0 proxy forwarded the response >> 7.0 the response reached the client >> 7.0 another client request for the same URL generated >> 7.0 client request reached the proxy, it is a HIT >> 7.0 proxy must compute Age header value, see math below >> >> Following RFC 2616: >> >> age_value = 0 (the cached response has no Age header) >> date_value = 0 (the cached response has Date set to 0) >> request_time = 0 (the proxy generated request at time 0) >> response_time = 7 (the proxy received response at time 7) >> now = 7 (the current time is 7) >> >> apparent_age = max(0, response_time - date_value) = 7 >> corrected_received_age = max(apparent_age, age_value) = 7 >> response_delay = response_time - request_time = 7 >> corrected_initial_age = corrected_received_age + response_delay = 14 >> resident_time = now - response_time = 0 >> current_age = corrected_initial_age + resident_time = 14 >> >> The true age is, of course, 7 and not 14. The above formulas just double true >> current age in the case of a network delay between the proxy and the origin >> server. The fixed formula (see below for the discussion) does not: >> >> current_age = now - min(date_value, request_time - age_value) = >> = 7 - max(0, 0 - 0) = 7 >> >> N.B. If the proxy computes Age header for misses and uses that as >> age_value when serving hits, the formulas yield the same result. >> >> >> The Problem >> ----------- >> >> RFC 2616 says: >> >> Because the request that resulted in the returned Age value must have >> been initiated prior to that Age value's generation, we can correct >> for delays imposed by the network by recording the time at which the >> request was initiated. Then, when an Age value is received, it MUST >> be interpreted relative to the time the request was initiated... >> So, we compute: >> >> corrected_initial_age = corrected_received_age >> + (now - request_time) >> >> I suspect the formula does not match the true intent of the RFC >> authors. I believe that corrected_initial_age formula counts >> server-to-client delays twice. It does that because the >> corrected_received_age component already accounts for one >> server-to-client delay. Here is an annotated definition from the RFC: >> >> corrected_received_age = max( >> now - date_value, # trust the clock (includes server-to-client delay!) >> age_value) # all-HTTP/1.1 paths (no server-to-client delay) >> >> I think it is possible to fix the corrected_initial_age formula to >> match the intent (note this is the *initial* not *received* age): >> >> corrected_initial_age = max( >> now - date_value, # trust the clock (includes delays) >> age_value + now - request_time) # trust Age, add network delays >> >> There is no need for corrected_received_age. >> >> >> Moreover, it looks ALL the formulas computing current_age go away with >> the above new corrected_initial_age definition as long as "now" is >> still defined as "the current time" (i.e., the time when current_age >> is calculated): >> >> current_age = corrected_initial_age >> >> So, we end up with a single formula for all cases and all times: >> >> current_age = max(now - date_value, age_value + now - request_time) = >> = now - min(date_value, request_time - age_value) >> >> It even has a clear physical meaning -- the min() part is the conservative >> estimate of object creation time. We could rewrite for clarity: >> >> creation_time = min(date_value, request_time - age_value); >> current_age = now - creation_time; >> >> >> Am I missing something important here? If I am right, and the current >> formulas count server-to-client delays twice, is it worth mentioning >> in the errata or elsewhere as a bug? Or should we insist that >> implementations use current_age calculation from the RFC anyway? >> >> Thank you, >> >> Alex. >> >> -- >> | HTTP performance - Web Polygraph benchmark >> www.measurement-factory.com | HTTP compliance+ - Co-Advisor test suite >> | all of the above - PolyBox appliance >> > > > -- > Mark Nottingham http://www.mnot.net/ > > -- Mark Nottingham http://www.mnot.net/
Received on Thursday, 4 March 2010 04:40:33 UTC