It wasn't like we had a discussion and all of a sudden decided this or that. When I needed to interop with Mark's service, he required the WS-Addressing support in the client, so I could implement it and be interoperable, or give up interop with his endpoint. As it turned out, the WS-Addressing support I put in was done last after all other stuff. I'm sorry if you wasted some time on this, but you also haven't been very vocal about your status yourself. Rich, Glenn, Mark and I have worked through lots of issues, but a lot of them come out in point to point issues. -- Chris On 06/11/06 09:12, "Peter G. Lane" <lane@mcs.anl.gov> wrote:
I don't think anybody colluded behind my back, don't get me wrong. I'm just trying to say that we should be more open about our discussion of what people are doing to achieve interop. For example, if a discussion between you and Chris had been held on the list about him implementing the headers in his client so that he could talk to your service, I would have picked up on this sooner and perhaps been saved from last-minute problems. So while I think it's likely largely a miscommunication, more open communication is perhaps part of it too.
I don't mean every little detail should have been discussed (Glenn and I are discussing a problem off list as we write), but summaries of problems and how it was overcome would be appropriate in case others run into the same problems. But perhaps everybody just implemented the soap headers without talking to anyone else, so I'll admit there's no guarantee this would have been caught sooner.
I'm also panicking since I've not been able to use anybody's service so far. So I apologize if I came off as a bit flippant.
Peter
Mark Morgan wrote:
I'm very sorry that you feel this way Peter and that you felt like you were forced to handicap your service, but as I have been trying to say, people have NOT been colluding behind your back. I am pretty sure that you misunderstood the conversation that you refer to and that at no time did anyone ever say that you couldn't use WS-Addressing EPRs. My recollection of that entire conversation was that it revolved primarily around whether or not people were required to use WS-Addressing EndpointReferenceTypes, not whether or not they were allowed to (modulo Chris Smith's one comment on how he would rather not implement WS-Addressing in his client).
Looking back over the email history of that conversation, it started with an email from you on 13 October 2006 which said, "Endpoints should be posted as EPRs, not URLs." , to which Glenn Wasson responded, "Not everyone needs an EPR to contact the service."
From these two comments it seems clear to me that this conversation is about whether or not people were required to post WS-Addressing EPRs to the WIKI, not about whether or not they were allowed to implement and use WS-Addressing for their endpoints.
Further, the Chris Smith email you refer to contains the following text, " Instead of:
https://aristotle.dreadnought.org:9090
You want:
<wsa:EndpointReference xmlns:wsa="http://www.w3.org/2005/08/addressing"> <wsa:Address>https://aristotle.dreadnought.org:9090</wsa:Address> </wsa:EndpointReference> "
Again, notice that this is about representation on the wiki, not about WS-Addressing as an allowed protocol. This message indicates the means by which the two representations (for degenerate EPRs) can be transformed back and forth. In that email, he does make the statement you refer to, "I'm guessing that yours will require me to set some SOAP header blocks based on the WS-Addressing SOAP binding. I had hoped to avoid this detail for now....", but this is about his client, not the interop fest as a whole.
From their the discussion turned largely into one of transport binding terminology, but the final word on the email list was a mail from you indicating that you would implement your service as a singleton that didn't require EPRs. Now that I read that, I can see how this could have happened. At that point, if someone was paying really close attention, they would have realized that you had just agreed to a course of action that was unneccessarily drastic. However, I think most people left it at that because your email seemed to declare that you had come to a decision internally and as such didn't need to discuss the issue further.
Again, I am truly sorry that you feel like you wasted your time turning your service into a singleton, but please realize that while this incident is unfortunate, it is not necessarily the case that anything underhanded or clandestine occurred. I believe that it was simple miscommunication that lead to this unfortunate accident.
-- Mark Morgan Research Scientist Department of Computer Science University of Virginia http://www.cs.virginia.edu mmm2a@virginia.edu (434) 982-2047
-----Original Message----- From: Peter G. Lane [mailto:lane@mcs.anl.gov] Sent: Monday, November 06, 2006 11:36 AM To: Mark Morgan Cc: ogsa-hpcp-wg@ggf.org Subject: Re: [ogsa-hpcp-wg] Globus Status
I'm a little frustrated because I thought I made it clear when I was pushing for EPRs as a requirement that we needed clients to support the WS-Adressing headers. But since it sounded as though people didn't want to bother with this, I capitulated and made my service use a singleton resource that didn't need an EPR, limiting it's capabilities and wasting time. So now I find that people have agreed off list to support such a scenario and I'm left with not being interoperable.
What I'm saying is that I don't have an objection to this in principle, but it would have been nice if someone would have clarified this earlier on the list when I was arguing the that my service would also need to add a ReferenceParameters element. I saw no discussion of people implementing the soap headers in clients so that they could interop with such services.
Peter
Well, personally, I believe that the discussion before revolved around whether or not people were required to post EPRs to the WIKI pages for degenerate EPRs where only the Address field had a value (i.e., where metadata and reference parameters were empty or null as allowed by spec.). Since such an EPR is identical in informational content to a play URL, people believed that they shouldn't be required to post EPRs. However, at no point did we decide to dissalow WS-Addressing of endpoint should an implementation choose to use them. Chris (and please correct me Chris if I mis-represent you) certainly expressed a desire to avoid implementing WS-Addressing on his client side, but has since done so in order to achieve interoperability with the GenesisII endpoint as our endpoint requires the ReferenceParameters header elements. Others in the HPC interop group, can you confirm or deny my interpretation of the early discussion? Is my endpoint out of spec., or am I correct in assuming that I am welcome to use WS-Addressing for my endpoint (i.e., to require ReferenceParameters in the headers.)?
-Mark
-- Mark Morgan Research Scientist Department of Computer Science University of Virginia http://www.cs.virginia.edu mmm2a@virginia.edu (434) 982-2047
-----Original Message----- From: Peter G. Lane [mailto:lane@mcs.anl.gov] Sent: Monday, November 06, 2006 11:09 AM To: Mark Morgan Cc: ogsa-hpcp-wg@ggf.org Subject: Re: [ogsa-hpcp-wg] Globus Status
This is what Chris said:
I'm guessing that yours will require me to set some SOAP header blocks based on the WS-Addressing SOAP binding. I had hoped to avoid this detail for now....
By this I was under the impression that nobody was going to build clients that added WS-Addressing SOAP headers. If that was
Mark Morgan wrote: the case,
then services shouldn't publish EPRs that have anything more than an address field.
Chris, did you change your mind? Was this only your point of view, or did you think that everyone was going to skip the SOAP headers as well?
At any rate, I should still be able to interop with those services that aren't expecting the SOAP headers. It just means I have to exclude at least the Genesis II implementation since I won't have enough time to implement the agreed upon version of WS-Addressing.
Peter
I think you misunderstood the result of that discussion. The end result was that services that didn't NEED EPRs, could use URLs. However, we didn't dissalow EPRs, only allowed the more degenerate case of them (the pure URL). -Mark
-- Mark Morgan Research Scientist Department of Computer Science University of Virginia http://www.cs.virginia.edu mmm2a@virginia.edu (434) 982-2047
> -----Original Message----- > From: ogsa-hpcp-wg-bounces@ogf.org > [mailto:ogsa-hpcp-wg-bounces@ogf.org] On Behalf Of Peter G. Lane > Sent: Monday, November 06, 2006 10:28 AM > To: ogsa-hpcp-wg@ggf.org > Subject: [ogsa-hpcp-wg] Globus Status > > I have the create and terminate operations implemented and working > locally. I will put up an endpoint sometime today, and by
Mark Morgan wrote: them will
> hopefully have the other operations implemented. Is there a WSDL > snipit or something of this "Show job output" operation, or does this > just mean that I have an FTP server running? > > Problems > -------- > I am unable to get responses from either of the .Net implementations > using my client. My SOAP message logging show that the client message > is going out, but nothing is ever received back. Are these services > still up? > > I tried using my client on the Platform service (service URL > https://plato.dreadnought.org), but I got an error saying that the > client couldn't find the correct CA. I did download and install the > CA cert and have debug output showing it was loaded, so I'm wondering > if anyone has had success with this service recently. If so, Chris, > could I get the public certificate of the server to help in my > debugging? > > After agreeing that we wouldn't use EPRs, I'm confused about the use > of them on the wiki. I did what others did and didn't bother > implementing support for the WS-Addressing SOAP headers. The Platform > EPR is trivial so that's not a problem. > The GenesisII EPR, though, is very non-trivial. That said, I see that > some clients have indeed talked to the GenesisII endpoint, so I'm > rather confused. Did I miss a discussion where it was decided that > full WS-Addressing support was now mandatory? > > Finally, I tried to use the CROWN endpoint too, but I get the > following error: > > Authentication failed [Caused by: Defective credential detected > [Caused by: [JGLOBUS-96] Certificate > "C=CN,ST=Beijing,L=Beijing,O=company,OU=department,CN=localhos > t" expired]] > > Peter >
-- ogsa-hpcp-wg mailing list ogsa-hpcp-wg@ogf.org http://www.ogf.org/mailman/listinfo/ogsa-hpcp-wg