*VidMid-VC Conference Call*
October 22, 2001

*Attendees*

Egon Verharen (chair) - SURFnet
Mary Fran Yafchak - SURA
Steve Olshansky - Internet2
Art Vandenberg - Georgia State
Jill Gemmill - ViDe/Alabama at Birmingham
Albert School - OARnet
Karen Krivaa - RADVISION
Michael Gettes - Georgetown
Renee Frost - Michigan/Internet2
Ted Hanss - Internet2
Ken Klingenstein - Colorado/Internet2
Tyler Johnson - UNC Chapel Hill
Samir Chatterjee - Claremont Graduate U.
Nadim El-Khoury - UNC Chapel Hill
John McNair - Tennessee
Ben Chinowsky (scribe) - Internet2

*Discussion*

The meeting opened with a short discussion of the October 2-5 Virtual Internet2 Member Meeting (VIMM); there was general agreement that overall the VIMM was a noteworthy success. [AI] All who have comments on specific successes and failures of the videoconferencing technologies and procedures used in the VIMM, will send them to Egon.

Renee gave a short overview of the NSF Middleware Initiative, launched on September 1. Jill and Renee noted that a major focus of VidMid's planned participation in this project is to "bless and wrap" existing applications, and make them interoperable. Watch http://www.nsf-middleware.org for developments.

Egon reported that he's made no progress on the VidMid-VC scenarios, and asked for help. [AI] Egon will send the latest versions of the scenarios to Art and Steve; Art will help write the scenarios and Steve will handle revision control.

The group discussed the overall middleware architecture for videoconferencing, taking as a starting point the Shibboleth-based architecture sketched by Ken in http://middleware.internet2.edu/video/draft-klingenstein-videoconf-arch-00.pdf. There was general agreement that VidMid-VoD is closer than VidMid-VC to being able to make use of the Shibboleth architecture, but that this model is worth pursuing for VidMid-VC as well. Tyler expressed a wish to get the Shibboleth team up to speed on ITU standards relevant to videoconferencing, and pointed the group to http://standard.pictel.com/ftp/avc-site/0103_Lau/0103_Lau.html for draft, and therefore free, versions of ITU standards.

Karen noted that H.225 Annex G has little support in the videoconferencing industry in general or from RADVISION in particular, and offered four reasons why this is the case: it's limited to the H.323 protocol and cannot serve systems with other signaling protocols such as SIP; it's not scalable, because of its technical architectural design; it's not widely supported by H.323 deployments -- vendors are not using it; and it offers limited services when compared with other possible solutions such as OSP. Karen asked the group to read up on OSP at http://www.openosp.org/whatis.htm. RADVISION now supports H.235 Annex D authentication and verification; in particular this is supported in a gatekeeper which will be on the market soon. The group discussed whether or not the VidMid-VC architecture should provide a Caller ID function; Tyler put this in the "nice but not essential" category, while Ken argued that "you can't design a service without designing for denial-of-service attacks" -- and in this case for "denial-of-privacy attacks" as well -- making something like Caller ID pretty important, though it need not be included in the very first stages of VidMid-VC's deployment. There was general agreement that network non-transparencies such as NATs and firewalls are going to be a big issue for VidMid-VC; [AI] Samir will send the list a short summary of approaches to coping with NATs and firewalls. There was also general agreement with Ken's suggestion that, as these issues are less of a problem in the the higher-education environment than elsewhere, the group can ignore then at first so as to concentrate on making progress in other areas.

With respect to directories, Tyler observed that the group already has a good idea of the attributes needed, and now needs to look more closely at the overall directory structure. Should the group create a directory-enabled gatekeeper? Should the video directory be separate from the enterprise directory? Tyler stressed keeping object class development closely guided by user scenarios, and suggested that the scenarios be incorporated into the object class documentation. There was general agreement that substantial time needs to be devoted to planning the VidMid-VC directory testbed, and that in particular VidMid-VC needs to a) decide whether to configure and ship servers or just publish object classes for each site to implement itself, and b) clarify requirements for testbed participants. Albert noted that he's made an OpenLDAP v2.x server and some IP addresses available for VidMid to use at vidmid.oar.net -- "I'll put in whatever anybody wants as long as it doesn't step on anybody." Albert also offered to maintain logs and provide documentation so this system can be more easily duplicated. How to incorporate RADVISION directory-enabled gatekeepers into the testbed is another issue; Karen asked VidMid to propose a timeline for her to take back to the developers at RADVISION, and Tyler urged the group to make this a priority, saying "RADVISION is offering to do a great thing for us here." Jill expressed interest in working on video directories within enterprise directories, using a shadow enterprise directory for testing, and there was general agreement that this approach should be pursued in parallel with the video-directories approach. Tyler said that his "ideal scenario is for people to implement on the enterprise directory with the caveat that the implementation may change," and Albert argued that while in the future it will be best to have all attributes in the enterprise directory, right now using a separate directory is useful for development and for convincing people to put its attributes in the enterprise directory. Multi-vendor interoperability will also need to be addressed sooner or later, despite MACE's apparent movement toward standardizing on iPlanet.

Another question confronting VidMid is, "to develop or to get the vendors to develop?" Egon suggested that the answer to this question should be the latter for H.323 but possibly the former for SIP. Tyler observed that "just white pages and the gatekeeper will enable many H.323 endpoints;" Jill noted that ViDe got interested in H.323 largely because it offered off-the-shelf components, and agreed that VidMid should not launch an H.323 development effort. Tyler noted Orit Levin's point that SIP is missing pieces in the area of capabilities exchange, and suggested that the group "make mistakes working with H.323, so we can start from a more advanced place with SIP." Samir has a SIP client that he can give away for use in VidMid, but it was developed on a commercial SIP stack for which his license has now expired. [AI] Samir will look into porting his SIP client to a standard open SIP stack. [AI] Samir will find out if and under what conditions dynamicsoft can make its SIP stack available to VidMid. Samir also noted that Microsoft is planning to give away a SIP client as part of Windows XP, but he's not sure what development tools will be provided for it. [AI] Karen will find out if and under what conditions RADVISION can make its SIP stack available to VidMid. [AI] Egon will contact Henning Schulzrinne for information on Columbia's SIP client. There was general agreement that developing a SIP client is a large project and that it would be overambitious for VidMid-VC to take this on at the outset. Jill and Tyler suggested that developing middleware interfaces for an existing SIP client might be a better project.

The group discussed key players missing from VidMid-VC, and what to do to get them involved. [AI] Karen will try to get RADVISION to identify and contact someone at Microsoft who can participate in VidMid-VC. [AI] Ted will keep trying to contact Microsoft and Polycom/Accord about participating in VidMid-VC. [AI] Jill and Egon will ask colleagues attending TeleCon to look for contacts for VidMid-VC. [AI] Egon will contact OpenH323 about participating in VidMid-VC. Samir and Egon noted Vovida's Red Hat-like business model; their SIP stack is open but they sell a maintenance agreement for it for $5,000/year. Samir listed the key players in SIP as RADVISION and dynamicsoft for the stack and PingTel for the client; he also mentioned WhitePine/CUSeeMe as a worth looking into, but according to Karen their reputation is poor. [AI] Tyler will look into Siemens's involvement in videoconferencing-related technologies, especially SIP. [AI] Egon will contact Philippe Galvez to check on the status of VRVS's participation in VidMid-VC. Tyler noted that Lucent's H.323 product line has been sold off and is being discontinued. [AI] Egon will review the vidmid-vc list to see who else should be contacted about participating.

The group agreed to make a priority of meeting face-to-face to thrash out the details of the overall videoconferencing middleware architecture and the VidMid-VC workplan, and in particular to plan testbed deployment. The meeting will take place November 29-30 at the Friday Center of the University of North Carolina at Chapel Hill; see http://www.fridaycenter.unc.edu/fc/index.html. More information about UNC Chapel Hill, including accommodations, is at http://www.unc.edu/depts/visitor/. VidMid-VC participants should RSVP to Steve ( steveo@luminagroup.com) as soon as possible.

Finally the group addressed its own week-to-week process. Unsurprisingly there was great interest in using videoconferencing for at least some of the VidMid-VC conference calls; however, it is not clear what videoconferencing capabilities VidMid-VC participants currently have in place. [AI] Steve will poll the list about using videoconferencing for future VidMid-VC calls. Renee noted that Internet2 can provide a bridge between videoconferencing and ordinary phone service, and Tyler offered to host the sessions at the UNC MCU and provide technical support -- "demands on my time are not an issue." There was a short discussion of whether VidMid-VC should use Groove or other document-sharing and revision-control tools; Steve noted that such tools are already being investigated by other groups within Internet2, advised VidMid to wait a few months to see what they come up with, and offered to perform an informal, manual revision-control function in the meantime. A password-protected area of the web site was also suggested. The group agreed on a call schedule of every other Monday from 11:00am-12:30pm EST (1600-1730 UTC), starting November 5th.

*Action Items*

[AI] 22-October-2001 - All who have comments on specific successes and failures of the videoconferencing technologies and procedures used in the VIMM, will send them to Egon.
[AI] 22-October-2001 - Egon will send the latest versions of the scenarios to Art and Steve; Art will help write the scenarios and Steve will handle revision control.
[AI] 22-October-2001 - Samir will send the list a short summary of approaches to coping with NATs and firewalls.
[AI] 22-October-2001 - Samir will look into porting his SIP client to a standard open SIP stack.
[AI] 22-October-2001 - Samir will find out if and under what conditions dynamicsoft can make its SIP stack available to VidMid.
[AI] 22-October-2001 - Karen will find out if and under what conditions RADVISION can make its SIP stack available to VidMid.
[AI] 22-October-2001 - Egon will contact Henning Schulzrinne for information on Columbia's SIP client.
[AI] 22-October-2001 - Karen will try to get RADVISION to identify and contact someone at Microsoft who can participate in VidMid-VC.
[AI] 22-October-2001 - Ted will keep trying to contact Microsoft and Polycom/Accord about participating in VidMid-VC.
[AI] 22-October-2001 - Jill and Egon will ask colleagues attending TeleCon to look for contacts for VidMid-VC.
[AI] 22-October-2001 - Egon will contact OpenH323 about participating in VidMid-VC.
[AI] 22-October-2001 - Tyler will look into Siemens's involvement in videoconferencing-related technologies, especially SIP.
[AI] 22-October-2001 - Egon will contact Philippe Galvez to check on the status of VRVS's participation in VidMid-VC.
[AI] 22-October-2001 - Egon will review the vidmid-vc list to see who else should be contacted about participating.
[AI] 22-October-2001 - Steve will poll the list about using videoconferencing for future VidMid-VC calls.
[AI] 24-September-2001 - Ken will send Philippe information on possible funding for integrating H.323 and SIP into VRVS.
[AI] 6-August-2001 - Samir will look into the availability and qualities of public SIP servers.