OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

ws-rx message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: RE: [ws-rx] AS need for ordered delivery?


Doug,
 
The whole point in having DAs represented is to be provide the capability to "view" the contract so that among a set of specific endpoints with different DA claims,  a client application can make a choice.
 
You made an excellent point. There is a very big distinction between what the contract is than requiring the RM protocol to define how to support the contract.
 
I am one of those people who is interested in addressing the former. This view should NOT be skewed/interpreted as being interested in changing the RM protocol or the definition of how to support this contract. We have hashed that the requirements on the RM protocol and its semantics and I do hope that there is general understanding that the tc is not going to address the latter.
 
Therefore, I find the ongoing discussion on removing DAs, etc not very useful. The TC has made a decision to include DA in the spec with the understanding that the goal is to specify the claim about DA, nothing more nothing less. 
 
Lets work with that.
 
Thanks,
 
--umit
 
 
 


From: Doug Davis [mailto:dug@us.ibm.com]
Sent: Thursday, Oct 27, 2005 7:36 PM
To: ws-rx@lists.oasis-open.org
Subject: RE: [ws-rx] AS need for ordered delivery?


+1

Short version: remove DAs entirely from the spec

Longer version: I think Duane's comparison with TCP is a very good one.  It illustrates how the upper layer treats the transport as a block box and that works well.  IMO, the RM layer should be a black box as well (for this discussion anyway  :-).  If a layer above TCP messes up the order of the packets before it actually reaches the TCP code then they're hosed.  If the layer above the RMS (meaning the AS) messes up the order of the soap messages before it reaches the RMS then they're hosed.  For that reason, the interactions/contract between the AS and the RMS are out of scope of this spec.  Likewise, for the most part, I think the interactions/contract between the RMD and the AD are out of scope too - with one exception, I do agree that the AS/RMS may want to know what the contract is - if for nothing else to know whether or not it wants to use that endpoint.  However, knowing what the contract is very different than requiring the RM protocol to define how to support that contract.  

That last sentence could lead one to believe that I might be ok with specifying the DA in Policy and adopting Anish's proposal for issue 6, which allows the RMS to specify the DA on the CreateSeq.  And to be honest I actually could go for that if I thought this entire DA discussion would end with that - but I doubt it would  :-)   I fear that any mention of DA in the spec would cause this discussion to continue to be rehashed over and over.  So, unless we can find the right text that would allow us to do what Anish is proposing w/o reopening this can of worms again my current leaning is to just remove it from the spec and let it be a problem for the Policy folks to work out.  After all, since it doesn't effect the protocol it just becomes a matter of advertising and negotiating the QoS levels of a service - which is a much bigger problem than our one little spec.

thanks,
-Doug




"Duane Nickull" <dnickull@adobe.com>

10/26/2005 07:53 PM

To
"Jacques Durand" <JDurand@us.fujitsu.com>
cc
<ws-rx@lists.oasis-open.org>
Subject
RE: [ws-rx] AS need for ordered delivery?





 
<SNIP/>
<JD> I notice though that TCP is designed - or implemented at least - so that the upper layer does not have to worry about packet reordering or numbering: so in effect TCP stacks interpret InOrder the same way we do so far in WS-RM: it is a black-box. Don't you prefer that to letting the upper layer "figuring it out" with packets numbers on its hand...?
<SNIP/>
Yes – that is exactly what I am implying.  A RMD is a black box that will be able to completely recreate the original stream as it was intended to be received based on a solid base WS-RX protocol.  What it does from that point on is discreet.  I am not in favor of specifying anything about the AD or any other upper layer.
 
My gut instinct is that DA’s should be removed all together from this work.
 
D


[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]