Interoperability Minutes 2009-05-11

From MemberWiki

Jump to: navigation, search

URL: http://www.openajax.org/member/wiki/Interoperability_Minutes_2009-05-11

Contents

Attendees

  • Matthias Hertel
  • Jon Ferraiolo, IBM
  • Javier Pedemonte, IBM
  • Eduardo Abe, ILOG/IBM
  • Howard Weingram, TIBCO
  • Bertrand Le Roy, Microsoft

Minutes

Topic: Not all samples working

Matthias: Some of the samples are not working

Javier: I haven't updated them yet. Will fix them today.

Matthias: Path not present error

Topic: Closing spec

Howard: Has anyone done a sync between JSDoc in open source and wiki?

Javier: Yes, I did it. Should be up to date.

Topic: Event names moved to separate chapter

Howard: I'm OK with that change

Topic: Sample code in Managed Hub Overview chapter

Jon: I updated the sample application in the Managed Hub Overview chapter. You all said the sample code should run, so I got it running and it is now checked into our open source project. I borrowed code from Adam's implementation of OpenAjax Widgets, particularly the XHR trick to load the inline client's HTML snippet from a file.

Howard: I want to look at it.

Topic: JSDoc documentation conventions

RESOLUTION: Move elsewhere, such as to an appendix

Howard: Can I go over it offline?

Jon: Yes

Howard: I need to do a full pass. I'll try to do it by Wednesday.

Topic: Walk through colored notes in Managed Hub APIs chapter

(Approved removing all colored notes except what is listed here)

Howard: Cross-site scripting note: change MUST to SHOULD

RESOLUTION: Agreed

Javier: onComplete has an error. I'll remove the brackets on the params.

Jon: OK, yes that was dumb.

(discussion about unsubscribe warning about unsubscribe being immediate)

Howard: Not sure it belongs here. I'm OK with it here, but I think are other cautions like this.

Matthias: Maybe include a tag like "Caution". Keep it here for easy reading.

Howard: Remove the sentence that references the SMash paper. We use the tunnel for things beyond what that paper says.

Jon: OK

Topic: APIs with no text

Matthias: Some of the APIs have no text. Just the JSDoc snippet. Is that OK?

Jon: I think so. The main description is in the JSDoc snippet. Why repeat it?

Matthias: Looks like something is missing.

Jon: How about "The JSDoc comments above fully describe this API."?

Matthias: Fine

Topic: Best practices chapter

(Regarding top comment on preserving URL parameters)

Howard: This is about how we reserve oah* parameters. Need to move this out of Best Practices into another part of the spec. The key thing here is that redirection needs to preserve URL params.

Jon: Not sure where this should go

Javier: I thought it was more of an implementation detail

Howard: We have containers. Put it there.

Javier: Yes, that would be good.

Jon: I'll work on it.

Howard: Add reference to Hub 1.0 spec

Jon: OK, I'll do it

Howard: Drop the comments about HUB10 and HUB20. Confusing

Matthias: Agreed

Jon: OK, I'll do it

Jon: I removed the section on read/write sections in publisherData

Howard: Main thought is that when dealing with iframes, you have to pass by value

Bertand: I understand and agree. Fine with me. But do you have a pointer to the old discussion?

Jon: I'll send a pointer.

Jon: How about the handle on disconnect section?

Howard: Can remove the 2nd bullet. We hide the iframe instead of remove it. If reconnect happens quickly , then container is happy.

Jon: OK, I'll consolidate the first bullet into previous sentence.

Jon: How about reconnect proposal?

Howard: I will review the spec and recommend whether we have enough information on reconnect or will send in a writeup with a detailed proposal

Jon: My question about what API does the client invoke to change the URL?

Howard: Private protocol within a container.

Topic: Redirect logic in open source

Javier: I commented out some of the code to make redirect work. Our implementation doesn't fit the scenario described in the paper exactly, so I think we are OK, but Howard, can you please take a look?

Howard: OK

Howard: I don't think that the code that is commented out exposes an attack surface.

Javier: Two levels up is hardcoded in our case. Precludes an intermediary attack.

Howard: Need to verify origin, not URL

Javier: Yes


Topic: Test suite and source code

Javier: I think test suite is far along. With source code, only think is the redirect issue.

Topic: refOrName

Jon: I sent email with 3 proposals. (1) Don't support refOrName, (2) Only support what is necessary for Hub10 compatibility, (3) Support it across the board

Howard: I'm in favor of (2) for backwards compatibility. Can always add it to 2.1. No need to do it across the board at this time.

Javier: I'd rather not have inconsistency.

Howard: I don't want to break backwards compability for Unmanaged Hub.

Howard: Only need to do it for Unmanaged.

Jon: Then there is a 2a and a 2b. 2a only restored Unmanaged Hub. 2b also allows refOrName on subscribe() callback across the board. 2a is better. Only have to change the code in one spot.

RESOLUTION: Only support refOrName for UnmanagedHub.

Topic: Upcoming phone calls

RESOLUTION: Freeze spec by Monday May 18

RESOLUTION: Try to be ready for Release Review announcement on Tuesday May 26

Personal tools