On 9/19/11 4:14 AM, Brett Porter wrote:
On 18/09/2011, at 4:13 AM, Olivier Lamy wrote:
Hi Folks,
With WAGON-348, the wagon-http based on Apache HttpClient include now
http connection pool mechanism.
Nice work :)
If no objections, I will commit this.
I'd say go ahead and commit.
One thing we should test - John had an ongoing battle with the wagons and
double uploading / checksum corruption. With the earlier versions of
wagon-http, that problem still existed.
I expect it'd be fixed by the combination of upgrading the httpclient and
http://jira.codehaus.org/browse/WAGON-347, but it'd be good if he could weigh
in. I'll see if I can reproduce on my end too.
IIRC, the big problem was that httpclient 3.x didn't use preemptive
authentication when you set a username/password. While the lightweight
implementation didn't either, it seemed to cache the fact that it used
authentication, and then preemptively use it on subsequent calls.
The effect of non-preemptive authentication was that you would PUT an
entire jar, zip, pom, whatever (some of which are quite large)...then
the server would reply with WWW-Authenticate, and you'd have to do it
all again, this time with credentials. I don't exactly remember how the
doubling of the checksum was resolved in the context of this problem, or
for sure that it was...but ISTR the checksums weren't the real problem
in the end, it was the upload timeouts on large files.
So, if httpclient 4.x has a better strategy for handling authentication,
I'd say we're good to go.
BTW, do any of the tests actually check PUTing a file and ensuring it
isn't sent twice?
- Brett
--
Brett Porter
[email protected]
http://brettporter.wordpress.com/
http://au.linkedin.com/in/brettporter
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
--
John Casey
Developer, PMC Chair - Apache Maven (http://maven.apache.org)
Blog: http://www.johnofalltrades.name/
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
