27-Apr-04 Meeting Minutes

["Yves Savourel" <ysavourel@translate.com>]

Hi all,
Here are the minutes.
-ys

Segmentation Subcommittee Meeting
April-27-2004

1- Roll Call:
Magnus Martikainen, Yves Savourel, John Reid, Eiju Akahane, Gérard Cattin des Bois.


2- Approval of Minutes from Last Meeting:
	The minutes were approved.


3- Open Action Items from previous meetings:

	* Tony: To look at what we did to XLIFF for making "embedded XLIFF" possible, so we could see for TMX and SRX have also such capabilities. Gérard thinks SRX was made to be embedded and requires no other action. He will confirm.
	+ Yves: To drive effort in OSCAR to make TMX a useable namespace if it's not already one.
		SRX was approved at last OSCAR SC meeting. It hasn't been made a 'formal' namespace, but Yves thinks there is not much to do to make it so. Same for TMX. He'll propose the modifications (if any) next OSCAR meeting. Gerard will look at URI with Arle and Tony.

	* All should address TM updating.
		Discussion was started in the mailing list

	* All should address workflow variations that can affect segmentation.
		Not much has been exchanged on that topic

	* Magnus will post modified definitions for both segment and segmentation.
		Done.

	* Andrzej will respond to latest email from from Magnus.
		Done.


4- Work in progress:

4-a) Discussion threads from last meeting:

4-a) i. Definition of Segment and Segmentation

Magnus posted new definitions. All like the fact it's rather generic: 'linguistic processing' is broad.
A discussion followed on the importance of fields (source, target, status, etc.) to define the segment.
John pointed out that we have already some fields that are singled out as they can be overwritten in <target>. Others are used to perform matches, etc. We may have to have a way to overwrite some attributes at the segment level too. This adds complexisty. Gerard noted that binding the definition to the 'fields' would be more flexible than trying to define the content of a segment.
All were in agreement that the latest definition was broad enough for now.


4-a) ii. Explicit XLIFF Segment representation vs. Multiply Converted XLIFF

Magnus summarized his views. we could use <group> and <trans-unit> to segment, but it would change the function of <group> and <trans-unit> are. John expressed concerns that changing the <trans-unit> makes it difficult to merge back the file without extra processing. We should have the capability to merge back the file at any point in the process.
Magnus noted that the fact XLIFF does not have currently a way of handling segments, it may push the filters developpers to use <trans-unit> for this. Yves agreed. Magnus pointed out that the filters should provide the most 'coarse' granilarity for the <trans-unit> (e.g. paragraph).
The case of translators needed to merge several <trans-unit> together was also discussed.
All agreed that there was a need for defining some kind of segment mechanism in XLIFF.


4.a) iii. Handling Segmentation Changes

Not enough time to discuss this.


4.b) Scenarios
4.c) Use Cases

All to post examples.


4.d) Implementation Options

All to post examples.


5. Any Other Business

None


Action Items:

	* Yves: To try to see what need to be done to make SRX and TMX a usable namespace.
	* Gerard: To check for a URI for namespace.
	* All: Post ideas for scenarios
	* All: Post ideas for representation