November 2005 - cddlm-wg

Notes from the meeting on 11/2/2005
by Milojicic, Dejan S 02 Nov '05

02 Nov '05

Attendees: Ayla Dantes De Souza, Stuart Schaefer, Jun Tatemura, Dejan Milojicic Note takers: Dejan Milojicic & Steve Loughran Interoperability update: Dejan spoke about interactions with Hiro, Cees, this will be in progress for a while,a s we learn from IETF how they do it. Jun: discussed his unit tests, independent from Deployment APIs and Component Model, Stuart uses Nunit Steve: Some interesting features of W3C testing; you know where tests are coming from; Every test has a unique ID. Stuart took the RDF team approach; take documetns and spit them out as valid output; so Stuart gets in CDL and oputs out the final compilation. So he then compares what input and output is. RDF: input and output. -Stuart and Ayla do this Proposed: test files for our content w/ input and output. Could machine generate the output and manually verify that things match. -test cases that we have. Inputs and output Take the parsed documents, the intermediate representation. Print it out to see what it looks like. Simple resolution still has a Dom Full resolution leads to a DAG and not a tree. Describing errors. We haven't defined how things fail, so cannot test for it. Agreed. There is nothing to stop additional implementation-specific testing on top. -if every test has a unique ID, then it's a matter of declaring extra metadata about each test, such as Give the result document and declare LAZY is not in scope. But we do need a standard way to represent a partially faild. SmartFrog testing? They exist, but if there is to be another impl. then those tests would need pulling out. Agreement Stuart wants more tangible things to send, so everyone is more consistent. -an example CDL for testing a system. -Here are the commands to issue. Steve to run through and revise test plan w/ more stuff. Nov16th. Jun to do CDL tests Steve plans to do an update on Nov 16th. Stuart to follow by December 1st.

1 0

Meeting tomorrow, 11/3, 6am PST
by Milojicic, Dejan S 02 Nov '05

02 Nov '05

Hi, My homework from the last meeting was to talk to Hiro about interoperability. I have done so and Hiro asked me to discuss this with Cees, whom I communicated with by email (see my earlier email today). I'll continue this investigation further with people who have pursued interoperability within IETF in the past. I also discussed with Hiro how we can continue our engagement with OGSA. We are scheduled to be on the OGSA call next Wednesday, 11/9. We will address the homework we had received while at F2F after GGF, and which we have addressed at the first meeting after GGF. The assignment for the meeting tomorrow was for everyone to think of the dependencies they may have from others so that we can start tracking progress we are making towards reference implementations. I would like to start tracking using GridForge trackers. Please use our usual phone numbers: +1 866-639-4732 (toll free while in US) Code 2362906 (followed by a '#' sign) +1 574-948-0379 (toll & international) Thanks, Dejan. > -----Original Message----- > From: Milojicic, Dejan S > Sent: Wednesday, October 26, 2005 6:49 AM > To: 'cddlm-wg(a)ggf.org' > Subject: Notes from the meeting > > Attendees: Stuart, Steve, Ayla, Dejan. > > Notes taker Dejan. > > We discussed how to demonstrate that something constitutes a > reference implementation. We suggested the following steps: > > 1. Individual components tests (every owner will contribute one). > > 2. Use cases: > > a) Single app > > b) Multi-tier, one machine > > c) Multi-tier, multi-machine > > 3. Optional: Grid computing apps (Blast, Globus) > > 4. Optional: Scale experiments > > Steve suggested that he will write a draft of the test > components by next week, a more complete suggestion in two weeks. > > Everyone will write a dependency list that they have on other > components so that we create a roadmap from now until next > GGF. One good example of dependencies is component tests that > should be available as early as possible so that reference > implementations can test their compatibility. We do not want > to track detailed plan of individual reference > implementations although we want to have some sense of the > overall progress we are making and whether we shall have > reference implementations completed for the next GGF. > > Dejan will talk to Hiro today to find out about the proofs > for interoperability and compatibility (if any). > > Thanks, > > Dejan.

4 4

Import resolution in CDL
by Steve Loughran 02 Nov '05

02 Nov '05

I am looking at relative import resolution. That is, if a document came from somewhere, then relative paths to resolve things are allowed. Who has implemented this already? Its an interesting feature of the specs is that we have left resolution undefined; import takes a string. Right now all I implement is the SmartFrog policy -all paths resolve to resources that must be loadable from the classpath -in secure mode, all .sf documents must be in signed jar files. If we allow arbitrary URLs to be resolved then -URLs must be resolved with the rights of the user doing the deployment. I must not be able to read .cdl docs to which I lack the granted rights. -We are going to need a way to secure deployment descriptors so that I dont download and deploy a subverted document. -Imagine I upload a set of cdl files using AddFile, and get back URLs to each of them. How can I put the URL to them into the CDL descriptor for importing? In the current HP resource-only-implementation, life is simpler as I require all the descriptors to go up as JAR files, then take a property to declare the extra set of jars to add to the classpath. But this will not interop with other apps. I will file this as an issue on the GGF trackers.

1 0

RE: Notes from GGF15
by Milojicic, Dejan S 02 Nov '05

02 Nov '05

:-) and the notes are attached this time. Dejan. > -----Original Message----- > From: Milojicic, Dejan S > Sent: Tuesday, November 01, 2005 4:35 PM > To: 'cddlm-wg(a)ggf.org' > Subject: Notes from GGF15 > > Hi Guys, > > Sorry for the delay, attached are consolidated notes from > GGF15 CDDLM meetings. Thanks to Kojo, Latha, and Steve for > taking notes at the three sessions. > > Thanks, > > Dejan.

1 0

Notes from GGF15
by Milojicic, Dejan S 02 Nov '05

02 Nov '05

Hi Guys, Sorry for the delay, attached are consolidated notes from GGF15 CDDLM meetings. Thanks to Kojo, Latha, and Steve for taking notes at the three sessions. Thanks, Dejan.

1 0

RE: How to demonstrate interoperability between reference implementations of a standard spec
by Milojicic, Dejan S 02 Nov '05

02 Nov '05

Many thanks Cees, I will pursue further research. Best regards, Dejan. > -----Original Message----- > From: Cees de Laat [mailto:delaat@science.uva.nl] > Sent: Tuesday, November 01, 2005 7:04 AM > To: Milojicic, Dejan S > Cc: Hiro Kishimoto > Subject: Re: How to demonstrate interoperability between > reference implementations of a standard spec > > Hi Dejan, > > In the IETF groups (Vendors) go out and make implementations > of the draft spec and organize their own interop events. Then > they write down an informational draft in which they note the > things that did need additional specification, errors, etc. > if any and they report basically the experiences. Based on > that report the working group in coordination with the GFSG > can conclude if you need another cycle to update the spec or > advance the spec to full standard. > > Some good examples can be found when you search for full > standards and see how they came to the end full standards track doc: > google: ietf full standard std64 > or std65 or any other. > > best regards, > Cees. > > At 11:20 AM -0700 10/28/05, Milojicic, Dejan S wrote: > >Hi Cees, > > > >The CDDLM-WG is currently working on four reference > implementations of > >our deployment specs and we are targeting demonstration of > >interoperability for the GGF17 time frame. There was no previous > >example of demonstrating interoperability within GGF. I have > discussed > >this with Hiro, the technical area director of the > management technical > >area to which CDDLM belongs. According to Hiro GGF draws its > practices > >after the IETF model. Because you are GGF's liaison with > IETF, we were > >wondering if you can assist us with how IETF approaches > demonstrating > >interoperability (use cases, documents, demos, etc.). Can you please > >help us with this? This is more so important because we will be a > >precedent in GGF and will pave the way for future interoperability > >demonstrations, therefore we need to do it the right way. > > > >Thanks, > > > >Dejan. > > > -- > http://www.science.uva.nl/~delaat/ >

1 0