If you are going to use the cookie in
such a manner, I would suggest architecting the system such that the cookie
is also reflected as a WSRP session. This raises it to the protocol level
and the releaseSessions() operation can be used to trigger the desired
cleanup.
Rich
07/14/2004 03:13 PM
To
cc
Subject
Re: [wsrp] What is the symetrical
method for initCookie?
1) producer will not know when to throw away the session cookie. In
the
absence of a releaseCookies operation, the only time a producer knows when
to kill the user session is when a logged in user's http session expires.
This, in my opinion, is a problem. We have seen cases where customers
bump
up the session expiry times to large periods. This happens mostly
in
non-shared machine environments and there are valid reasons for it. In
such situations, if the system just depends on the session expiry times
and
not the user logging out, the webserver runs out of memory due to the
orphaned sessions.
Essentially, we need someway for the consumer to tell the producer that
a
user has logged out. It doesn't really matter that it is a releaseCookie
call or just a userLogout call -- something that can help the producer
identify and cleanup the unneeded resources. This is needed with
or
without the initCookie i.e. even if a system doesn't set cookies but uses
url re-writing to maintain sessions. In any non-trivial system, producers
will have resources tied up to a logged in user. As Ricky wrote,
this
opens the system up to denial of service attacks, even without the attacks
the system can run out of memory if the expiry times are large as I stated
above. This and security were the main reasons why logouts were invented
in the first place.
The main question is whether WSRP which is an application protocol be
handling this or should this be handled by something like WS-Security.
What are your thoughts on that?
"Rich Thompson"
< To:
m>
cc: (bcc: Khurram Mahmood/PeopleSoft)
Subject: Re: [wsrp] What is the symetrical method for
initCookie?
07/14/2004 11:37
AM
When this was discussed, it was decided that a releaseCookies() was not
needed since 1) the Producer may throw cookies away at any time it desires
and 2) initCookie() was only placed into the protocol due to the unique
initialization needs of clustered servers (and viewed by most as a
pollution of the protocol!).
Rich
07/14/2004 02:21 PM
To
cc
Subject
[wsrp] What
is the symetrical method
for initCookie?
It seems that unless there is a method like "releaseCookie" won't
the
producer be open to DoS attack, or more likely just running out of
resources on a busy server.
Thanks
To unsubscribe from this mailing list (and be removed from the roster of
the OASIS TC), go to
http://www.oasis-open.org/apps/org/workgroup/wsrp/members/leave_workgroup.php
.
To unsubscribe from this mailing list (and be removed from the roster of
the OASIS TC), go to http://www.oasis-open.org/apps/org/workgroup/wsrp/members/leave_workgroup.php.
