axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aleksander Slominski <as...@cs.indiana.edu>
Subject Re: [Axis2] Proposal - DOM Implementation for Axis2
Date Wed, 18 May 2005 00:36:20 GMT
Sanjiva Weerawarana wrote:

>On Mon, 2005-05-16 at 13:22 -0500, Aleksander Slominski wrote:
>  
>
>>hi,
>>
>>doesn't it mean that XML will be parsed twice or events replayed 
>>(PullParser pp = messageContext.getSOAPEnvelope().getPullParser())  and 
>>original OM tree that was first built is then discarded (first built by 
>>StAXSOAPModelBuilder but without DOM then again by StAXSOAPModelBuilder 
>>but with DOM)?
>>    
>>
>
>Yep that's what it means.
>
>This is unavoidable .. its a chicken-n-egg situation. We only want to
>pay the price of building a DOM (instead of an OM) when WS-Sec (or
>someone else) wants it. Now, if the entire system has been set up with
>WS-Sec turned on then yes we can build that straight away and no
>problem. However, often WS-Sec will be on for some number of services,
>but not for all. In order to identify the service you need to read the
>headers (maybe .. at least if ws-addr is in use). Thus the chicken comes
>before the egg: have to build an OM so we can find the service so that
>we can decide that we need a DOM (the egg). Oops.
>  
>
i have yet to see example of service that will accept secure *and* non 
secure messages on the same transport endpoint
(HTTP has multiple paths, JMS have multiple queues, etc etc for any 
transport)
and typically it is done by deploying two endpoints (it could be done on 
one endpoint but why to do it?)

>The proposal Ruchith made solves that problem by allowing you to.
>switch .. albeit at a cost of duplicating the part of the tree that has
>already been built (by piping that thru StaX). 
>
AXIS2 is exposing services and if deployer choose to have unique HTTP 
endpoint (http://host/service/Foo and 
http://host/service/FooWithMessageSecurity) why not use this to AXIS2 
advantage -
this would say something about flexibility of internal 
modules/handlers/deployment  architecture ...

>BTW I'm very impressed
>with our OM design that it allowed us to easily make this work :-).
>  
>
nice it can do it but i would tell users to avoid putting AXIS2 runtime 
in situations like that ...

>Plus, in any case, in Sri Lanka we know damned well the chicken came
>before the egg anyway; how else could the egg have come into existence
>at all?
>  
>
right just the first chicken may not meet standards we have for today 
chickens ...

alek

-- 
The best way to predict the future is to invent it - Alan Kay


Mime
View raw message