[
https://issues.apache.org/jira/browse/HTTPCLIENT-1163?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13213945#comment-13213945
]
Jon Moore commented on HTTPCLIENT-1163:
---------------------------------------
I think the underlying issue is actually that the CachingHttpClient does not
work with a ContentEncodingHttpClient injected as the backend client, due to
the modification (uncompressing) of the response body done by the
ContentEncodingHttpClient; this issue is also reported as HTTPCLIENT-1164 and
HTTPCLIENT-1167. As Joe demonstrated, the Vary processing is not actually the
problem; the problem is that we have a cache miss due to the un-gzipped
response having a length that does not match its Content-Length header.
As such, I'll close this particular issue as a duplicate and address it with
the other two.
> Incorrect processing of Vary: HTTP header of cacheable server response
> ----------------------------------------------------------------------
>
> Key: HTTPCLIENT-1163
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-1163
> Project: HttpComponents HttpClient
> Issue Type: Bug
> Components: Cache
> Affects Versions: Snapshot
> Reporter: Manish Tripathi
> Assignee: Jon Moore
> Labels: patch
>
> Vary header from the server's response is not processed correctly by the
> cache when prepending the list of variables to the URL.
> Example (unrelated headers removed from the requests/responses for clarity).
> -> client sends:
> GET http://s.ytimg.com/yt/cssbin/www-guide-vflgAVfxE.css HTTP/1.1
> Accept-Encoding: gzip, deflate
> Host: s.ytimg.com
> -> server responds:
> HTTP/1.1 200 OK
> Vary: Accept-Encoding
> Current implementation produces the following variant URI to be stored in the
> cache:
> {Accept-Encoding=}http://s.ytimg.com/yt/cssbin/www-guide-vflgAVfxE.css
> Notice how the actual value of Accept-Encoding header from the original
> request is NOT prepended.
> The correct variant URI should be:
> {Accept-Encoding=gzip%2Cdeflate}http://s.ytimg.com/yt/cssbin/www-guide-vflgAVfxE.css
> The cause of the problem:
> org.apache.http.impl.client.cache.CachingHttpClient.java, method
> callBackend(..)
> uses the original version of HttpRequest without headers added by
> DefaultHttpClient and/or ContentEncodingHttpClient implementation.
> Proposed fix:
> Replace the following lines of
> org.apache.http.impl.client.cache.CachingHttpClient.java:
> HttpResponse backendResponse = backend.execute(target, request,
> context);
> backendResponse.addHeader("Via", generateViaHeader(backendResponse));
> return handleBackendResponse(target, request, requestDate,
> getCurrentDate(),
> backendResponse);
> with the lines:
> if (context==null) context=new BasicHttpContext();
> HttpResponse backendResponse = backend.execute(target, request,
> context);
> backendResponse.addHeader("Via", generateViaHeader(backendResponse));
> HttpRequest
> actualRequest=(HttpRequest)context.getAttribute(ExecutionContext.HTTP_REQUEST);
> return handleBackendResponse(target, actualRequest, requestDate,
> getCurrentDate(),
> backendResponse);
> After the corrections above, further experiments show that variant cache
> entries are still not being processed correctly. In the example above, the
> cache entry will be stored with the key
> "{Accept-Encoding=gzip%2Cdeflate}http://s.ytimg.com/yt/cssbin/www-guide-vflgAVfxE.css";,
> but lookups will be performed using the key
> "http://s.ytimg.com/yt/cssbin/www-guide-vflgAVfxE.css";, because obviously the
> client does not know that this particular cache entry has variants at the
> time of request.
> So maybe the correct fix would be to remove this
> "{Accept-Encoding=gzip%2Cdeflate}" prefix from the key completely when
> storing it in the cache?
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
