0000912: Put proxy authentication back in HTTPSocket

(0001234 - 101 - 101 - 101 - 101 - 101 - 101)
Hans-Martin
03-08-05 13:12

Monticello also includes a change which removes the proxy authentication, so it should be fixed, too.

(0001235 - 213 - 213 - 213 - 213 - 213 - 213)
dfaught
03-08-05 16:57

I know that the Monticello overlap was being worked on a while back, and thought it might be fixed by now. It is very possible to have both proxy authentication and server zone authentication in a single request.

(0001398 - 74 - 74 - 74 - 74 - 74 - 74)
dfaught
04-20-05 16:02

The newer attached file fixes this again after change 6651ExternalCleanup.

(0001461 - 596 - 596 - 596 - 596 - 596 - 596)
dfaught
05-05-05 18:36

The current version of Monticello on SqueakMap does indeed still contain a change that removes proxy authentication. I vaguely remember Avi saying something about being backward compatible so he couldn't remove the method from the changeSet, even though it wasn't really needed any more. Two points here: 1. proxy authentication is in fact in the current image and Monticello (mainly) is not. This is a bug fix to something in the image and should be put in. 2. Someone needs to figure out how to update Monticello so that it also includes this fix or doesn't include the method in question.

(0002933 - 159 - 159 - 159 - 159 - 159 - 159)
KenCausey
10-20-05 21:00

Please review the current code in 3.9a-6693md5 (or later) it appears to me that proxy support is currently included and so this issue can be closed. Accurate?

(0002952 - 301 - 325 - 325 - 325 - 325 - 325)
Hans-Martin
10-22-05 11:37

The problem is still there.
It's in the method HTTPSocket class>>httpGetDocument:args:accept:request:
In the request headers, it should include the contents of HTTPProxyCredentials when going through a proxy.
Since the header is ignored when not going through a proxy, it can be added in all cases.

(0002975 - 408 - 420 - 420 - 420 - 420 - 420)
KenCausey
10-25-05 17:30

David and Hans-Martin,

I would really appreciate it if you could look at the HTTPProxyCredentials changeset I've attached to this issue. This is my attempt to adapt David's changes to what is currently in Squeak3.9a and I've made the use of the HTTPProxyCredentials contingent on actual use of the proxy server. If you can test this (this changeset is against 3.9a-6693md5) I would really appreciate it.

(0002987 - 162 - 168 - 168 - 168 - 168 - 168)
dfaught
10-26-05 17:15

Ken,
The reformatting somewhat obfuscates the fact that this is basically a one line change, but yes, it works just fine. Thank you for your work on this issue!

(0004310 - 142 - 148 - 148 - 148 - 148 - 148)
nm
03-02-06 23:37

I see this issue is still waiting for feedback... the fix is working. Will it be integrated in 3.9 ?
I'm submitting the fix against 3.9a-7004

(0004311 - 292 - 292 - 292 - 292 - 292 - 292)
KenCausey
03-02-06 23:45

The hold-up is that HTTPSocket is based on OldSocket which is destined for the chopping block, which may also be the destiny of HTTPSocket. The I/O Stewards team is still getting it's ducks in a row and I hope we can make some progress on this issue and related issues in the next few weeks.

(0005031 - 326 - 350 - 350 - 350 - 350 - 350)
KenCausey
05-23-06 19:06

At this point it seems unlikely that we are going to do anything about OldSocket or changing HTTPSocket for 3.9.

So we need to finish up this issue. Can someone please try this in the most recent 3.9beta and confirm that it still does the desired job? I'm not behind a proxy server so I can't reliably test it.

Thanks.

Issue History
Date Modified	Username	Field	Change
03-02-05 21:14	dfaught	New Issue
03-02-05 21:14	dfaught	File Added: HTTPAuthAgain-daf.1.cs
03-08-05 13:12	Hans-Martin	Note Added: 0001234
03-08-05 16:57	dfaught	Note Added: 0001235
04-20-05 16:01	dfaught	File Added: HTTPAuthAgain2-daf.1.cs
04-20-05 16:02	dfaught	Note Added: 0001398
05-05-05 18:36	dfaught	Note Added: 0001461
10-20-05 20:59	KenCausey	Status	new => assigned
10-20-05 20:59	KenCausey	Assigned To	=> KenCausey
10-20-05 21:00	KenCausey	Note Added: 0002933
10-20-05 21:00	KenCausey	Status	assigned => feedback
10-22-05 11:37	Hans-Martin	Note Added: 0002952
10-25-05 17:28	KenCausey	File Added: HTTPProxyCredentials.1.cs.gz
10-25-05 17:30	KenCausey	Note Added: 0002975
10-26-05 17:15	dfaught	Note Added: 0002987
03-02-06 23:37	nm	Note Added: 0004310
03-02-06 23:38	nm	File Added: ProxyAuthFix.1.cs.gz
03-02-06 23:41	nm	Issue Monitored: nm
03-02-06 23:45	KenCausey	Note Added: 0004311
03-02-06 23:49	nm	Issue End Monitor: nm
05-23-06 19:06	KenCausey	Note Added: 0005031
01-23-09 19:52	KenCausey	Relationship added	has duplicate 0007193

Notes
(0001234 - 101 - 101 - 101 - 101 - 101 - 101) Hans-Martin 03-08-05 13:12	Monticello also includes a change which removes the proxy authentication, so it should be fixed, too.

(0001235 - 213 - 213 - 213 - 213 - 213 - 213) dfaught 03-08-05 16:57	I know that the Monticello overlap was being worked on a while back, and thought it might be fixed by now. It is very possible to have both proxy authentication and server zone authentication in a single request.

(0001398 - 74 - 74 - 74 - 74 - 74 - 74) dfaught 04-20-05 16:02	The newer attached file fixes this again after change 6651ExternalCleanup.

(0001461 - 596 - 596 - 596 - 596 - 596 - 596) dfaught 05-05-05 18:36	The current version of Monticello on SqueakMap does indeed still contain a change that removes proxy authentication. I vaguely remember Avi saying something about being backward compatible so he couldn't remove the method from the changeSet, even though it wasn't really needed any more. Two points here: 1. proxy authentication is in fact in the current image and Monticello (mainly) is not. This is a bug fix to something in the image and should be put in. 2. Someone needs to figure out how to update Monticello so that it also includes this fix or doesn't include the method in question.

(0002933 - 159 - 159 - 159 - 159 - 159 - 159) KenCausey 10-20-05 21:00	Please review the current code in 3.9a-6693md5 (or later) it appears to me that proxy support is currently included and so this issue can be closed. Accurate?

(0002952 - 301 - 325 - 325 - 325 - 325 - 325) Hans-Martin 10-22-05 11:37	The problem is still there. It's in the method HTTPSocket class>>httpGetDocument:args:accept:request: In the request headers, it should include the contents of HTTPProxyCredentials when going through a proxy. Since the header is ignored when not going through a proxy, it can be added in all cases.

(0002975 - 408 - 420 - 420 - 420 - 420 - 420) KenCausey 10-25-05 17:30	David and Hans-Martin, I would really appreciate it if you could look at the HTTPProxyCredentials changeset I've attached to this issue. This is my attempt to adapt David's changes to what is currently in Squeak3.9a and I've made the use of the HTTPProxyCredentials contingent on actual use of the proxy server. If you can test this (this changeset is against 3.9a-6693md5) I would really appreciate it.

(0002987 - 162 - 168 - 168 - 168 - 168 - 168) dfaught 10-26-05 17:15	Ken, The reformatting somewhat obfuscates the fact that this is basically a one line change, but yes, it works just fine. Thank you for your work on this issue!

(0004310 - 142 - 148 - 148 - 148 - 148 - 148) nm 03-02-06 23:37	I see this issue is still waiting for feedback... the fix is working. Will it be integrated in 3.9 ? I'm submitting the fix against 3.9a-7004

(0004311 - 292 - 292 - 292 - 292 - 292 - 292) KenCausey 03-02-06 23:45	The hold-up is that HTTPSocket is based on OldSocket which is destined for the chopping block, which may also be the destiny of HTTPSocket. The I/O Stewards team is still getting it's ducks in a row and I hope we can make some progress on this issue and related issues in the next few weeks.

(0005031 - 326 - 350 - 350 - 350 - 350 - 350) KenCausey 05-23-06 19:06	At this point it seems unlikely that we are going to do anything about OldSocket or changing HTTPSocket for 3.9. So we need to finish up this issue. Can someone please try this in the most recent 3.9beta and confirm that it still does the desired job? I'm not behind a proxy server so I can't reliably test it. Thanks.