Thanks for your comments. They are very much appreciated.
Responses to comments (below).
a)
The initialisation parameter part is a little confused. Where I say uri
parameters in 3.2 I mean the parameters to the action uri, not the request
uri. Parameters in the action uri may be user specific, but they are
consistent across the proxylet each time the proxylet is used by the user.
Currently, if implemented as written there would be an instance of the
proxylet for each action uri that is used.
Parameters like user-id that you mention would be present in the session
if you had been through an authentication process, otherwise they could be
present in the request uri, or the cookies. Theres no problem in initialising
against user specific parameters, there may be inefficiencies related to
shared resources.
I dont really like the action uri mechanism in IRML. The fact that we are
mixing up url encoding and xml to format data. I think it would be better if
be had a xml tag for the parameters to pass to the proxylet. I also dont like
the specification of hostname there.
The action should look like: -
<action service="thundercrack/personalize" >
<parameter name="user-name" value="andy" application-point="usage"/>
<parameter name="language" value="English" application-point="init" />
</action>
This allows
i. the omml to specifiy what callout mechanisms are available, and to
choose
the best host to do it on. The irml author need not know about the particular
deployment scenario, just that the named service is present. Its likely in
the CDN deployment scenario that you wont know which cluster you are being
served by nevermind which servers are present.
ii. The parameters can be split into initialisation parameters and per
action
parameters allowing for more efficient resource sharing.
b+c) I'll make it more consistant.
d) Client means next box down the line.
e) yes.
Regards,
Andy
On Monday 17 September 2001 7:34 pm, you wrote:
Andrew,
a.
In section 3.2 Initialisation,
it is mentioned that "parameters" mentioned in URI can be passed
here. I just want to know, if we use proxylet which doesn't implement
SingleThreadedProxylet, and parameters mentioned are "user-specific", such
as "user-id=staff.foo", is it ok to pass such parameters. Or we can't
specify such parameters in uri.
whether, I misunderstood you.
b.
In 6.7.3, Description should be about "Protocol Version".
c.
In 6.7.2 and 6.7.3, we use two APIs to determine protocol name and
protocol version in case of ProxyletRequest, but in case of Proxylet
response( 6.8.2), we use only one API to gather the same information. Why
this inconsistency?
d. In section 6.7.10, does "client" mean "end user" or "immediate box
down ".
3. In section 6.8.5, "Description" is little confusing. Also, returned
IP address will be of "origin server" or one level up box.
Rajnish