[byteio-wg] Telcon today + Minutes from GGF

[neil p chue hong]

Hi All,

there is a regular telcon today at 14:00 UTC, 15:00 BST, 10:00 EST
Dialin numbers:
UK - 0870 240 7821
USA (east) - 718 354 1169
Germany - 06950070811
International: +442078193600
Passcode: 14712935# 

The main topic for the agenda is reviewing the discussion at GGF and
building the issue list for the documents.

cheers,
neil

ByteIO Meeting minutes @ GGF15
===========================

Participants:
	Mark Morgan (Chair, dial-in), UoV
	Neil Chue Hong (Chair), EPCC
	Malcolm Atkinson, NeSC
	Michel Drescher (note taker), FLE
	Donal Fellows, UoM
	Amy Krause, EPCC
	Stephen Davey (note taker), NeSC
	Dave Berry, NeSC
	Pascal Kleijer, NEC

- Neil introduces the GGF Intellectual Property notice

- Neil gives an introduction to ByteIO
	* slides are available on the project's Gridforge project page
	* bi-weekly phone conferences

- Neil gives an overview of the ByteIO use cases
	* File use case
		+ flat file data availability
		+ spans into three "sub use cases"
			> Grid Aware ByteIO Access
			   (read/write to a remote "grid" file)
			> NFS / CIFS Server
			   (connecting legacy applications)
			> stream oriented access
			   ()
	* Data Access use case
		+ Asynchronous use case
		+ Simple and complex Insert
		+ Copying
	* Visualisation Use Case
		+ defined as part of the SAGA-RG

ACTION ITEM: Neil to study the latest straw man of the SAGA-RG

- Neil presents the ByteIO interfaces (using marks slides)

Block interface:
-------------------

Question Malcolm: What's the "I" stand for?
Answer Mark: Comes from a naming convention from Java Programmers
Malcolm: I is misleading as it could have meant "Immediate" as opposed to
"deferred", also it is clear that interfaces are currently specified
Solution: Delete the I

Question Dave Berry: why is the interface stridable?
Mark: Based on the comments back at GGF-14 (SAGA-RG)

Malcolm: Start offset is easy for files - how is it for entities in memory?
Mark: This is an implementation detail

Question Donal: Why is stride unsigned? You may want to stride backwards...
Answer Mark: No particular reason, happy to change

Question: return type of read is byte array. Does the operation allocate
that memory, or is a buffer to be provided?
Answer: The specification provides a port type, so this is left as an
implementation detail

Question: Why not join append and truncAppend?
Answer (group): the resulting operation would not have been transactional
atomic.

Streamable interface:
----------------------------

Question: Why is there an offset in the seekRead / seekWrite?
Answer: Where implementations cannot provide an offsetable read/write, an
appropriate fault should be returned

Question: Why do you not provide for read in the random interface for
negative offsets (i.e. from the end of the file)?
Answer: (answer missed)

Question: Does seekRead(streamable) ensure the number of bytes?
Answer: None of the read requests do NOT ensure the requested number of
bytes

ACTION ITEM FOR NEIL: To provide an example how to efficiently read from the
end from a log file that is permanently appended using ByteIO interface


RandomByteIO properties
---------------------------------

ACTION MARK: change the minOcurs of Modification Time and Access TIme to "0"

ACTION MARK: clarify timestamps remarks with respect to the clock sync
issues if hosts wwhen creating a resource (i.e. file)

ACTION MARK: change ModificationTime and AccessTIme to MAY rather SHOULD


StreamableByyteIO properties
--------------------------------------

ACTION MARK: To add a boolean seekable property

Question: EndOfStream does not know the Unix style behaviour of
resurrection. Is this intended?
Answer: This is by coincidence, and should not be changed


Lifetime Management
----------------------------

- Look at the Rendering document
- Creation: 
	* out of scope  but closely related
	* for streamable byteIO, users should be able to identify both an
open mode and an access mode
- Deletion:
	* Destruction is very much in scope
	* Different semantics for Streamable and Random byteIO

ACTION MARK/NEIL: To add  a clarification about Concurrency in ByteIO

Faults and Failures
-------------------------

There was discussion which was not captured.



-- 
Neil P Chue Hong            | T: [+44] (0)131 650 5957
Project Manager, EPCC       | F: [+44] (0)131 650 6555
Rm 2409, JCMB, Mayfield Rd. | E: N.ChueHong at epcc.ed.ac.uk
Edinburgh, EH9 3JZ, UK      | W: http://www.epcc.ed.ac.uk      

BT MeetMe: http://tinyurl.com/8mwhd - Code: 14712935#  

  "A film is like a battleground. It's love, hate, action,
   violence, death - in a word, emotion." - Sam Fuller