Data Transport Research Group
GGF7, Tokyo, 6th March 2003

Attendees: ~55

Coordination with other groups (Grid High Performance Networking,
Replication Networking and IETF):

- there were no volunteers to coordinate interaction with these
  groups.

Milestones and lifetime of group:

- the group charter will need to be amended to give it a definite
  lifetime. It was suggested that we should use the timescale of the
  milestones to define the end point of the group

- each milestone being discussed at the meeting was asked to suggest
  timescales for their deliverables

Survey of Transport Protocols over than TCP:

- Bill Allcock presented on behalf of Jason Leigh
- Are doing a survey of open literature to produce a document with one
  page per protocol.
- A draft list of protocols had been finished.
- Discussion over whether survey should cover implementations or
  focus strictly on protocols. Agreed to include implementations
  where appropriate and people willing to contribute.
- Draft list of protocols to be posted to the mailing
  list. Other groups should contribute other protocols they're
  aware of.
- Bill and Jason would produce and circulate a template for an
  entry. This would allow other groups to contribute, even
  partially, to the survey.
- Iterate this document to GGF8 and produce a draft for
  discussion. Submit as an Informational document at GGF9.

Recommendations for Standard Operations at remote sites:

- Reagan Moore presented results of an initial survey of five
  existing implementations used in current Grids: SRB, CHIRP,
  DCCP, RFIO, RIO/MPI-IO
- Want to identify a common subset and also find out what
  communities use "other" operations.
- Require participation from active Grid communities to say
  why a particular operation should be a standard grid operation.
- Initial list to be published to mailing list to get feedback
- Try to submit a document for GGF8 if enough participation.

Extensions to HTTP for the Grid:

- Andy McNab presented some work looking at using HTTP(S) for Grid
  data transport.  

- Could lead to a number of different documents: 
-- Informational (what can be acheived, performance vs other protocols)
-- Best Practice (how existing standards should be used to achieve
   particular performance / functionality) 
-- Standards (which bits of standards need to go from MAY to SHOULD/MUST) 
- Unclear how much interest there is in this area at GGF.  
- relying on HTTP extension headers might be a problem (see the IESG
  note page 1 of RFC 2774). SOAP could be more appropriate.
- Andy to bring a draft of a possible document to GGF8.

Protocols improvements to GridFTP v1.0

- Igor Mandrichenko presented some experiences of using GridFTP v1.0
- Goals: summarise implementation experiences; recommend solutions;
  develop next version of standard; implement new standard
- likely to become either revival of GridFTP WG or new spin-of WG
- Bill Allcock would present a list of known problems with GridFTP
  v1.0 from the Globus Project. Other people should contribute
  additions to this list.
- Timescale: find interested parties between GGF7/GGF8, combine
  opinions and form scope, produce a charter for GGF8.

Requirements for Asynchronous and Heterogenous Data Transport

- Michael Welzl presented a "Specification of a Network
  Adaptation Layer for the Grid"
- Grid applications have special QoS requirements / traffic
  properties => does this require special architecture?
- Should this be a RG? WG? Document? Need to see what interest
  there is in this area before GGF8.
- This activity might be better suited to an IRTF RG. Once a proposal
  has been put together there, the suitability of this approach to
  Grids could be investigated within this RG.

Minutes by Neil Chue Hong, EPCC <N.ChueHong@epcc.ed.ac.uk>