May-11 Meeting Minutes (with corrections)

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

Here is an updated version of the minutes.

(Thanks for the corrections Magnus.)

Cheers
-yves

Segmentation Subcommittee Meeting
May-11-2004 (rev. 2)


1- Roll Call:
Present: Magnus, Yves, John, Eiju, Andrzej.
Appologies: Gerard, Christian, Tony.

2- Approval of Minutes from Last Meeting:
	John move to accept the minutes, Yves seconded. Minutes were approved.


3- Open Action Items from previous meetings:

	* Yves: To try to see what need to be done to make SRX and TMX a usable namespace.
		Yves reported that both SRX and TMX had URI defined for namespace and even examples, but where (so far) meant to be used as a 'document' embeded inside another. Andrzej added his experience with SRX: it can be used be used to define segmentation anywhere (it's not build explicitlty for TMX). Magnus said it may be cases when re-using SRX rules from TMX to XLIFF where we could have some issues and need to re-write partially the regular expressions.

	* Gerard: To check for a URI for namespace.
		Gerard sent his appologies. Yves: yes there are.

	* All: Post ideas for scenarios
		None new was posted.

	* All: Post ideas for representation
		Yves posted one representation.


4- Work in progress:

4.a) Reach conclusion regarding Explicit XLIFF Segment representation vs. Multiply Converted XLIFF

Magnus and Andrzej met in Dublin last week and talked about it. The conclusion is that multiply converted XLIFF may be ok in some case, but not always. So it seems that we need an explicit XLIFF segment representation. A general consensus for this was there.


4.b) Reach conclusion regarding the need for representing Segementation in XLIFF

A general consensus that there for this too.


4.c) Discuss Scenarios and Use Cases

Magnus listed from the top of his head the few cases we had so far:
- Using XLIFF as a generic exchange (may not need segmentation)
- Using XLIFF during a localization project (probably need segmentation)
- Existing XLIFF files based on linguitic aspects that is adopting segmentation of an existing XLIFF file for use with existing linguistic assets.
  This case further splits into two cases:
	a) the XLIFF file is "source only"
	b) the XLIFF file has both source and target content
- Filtering: trying to keep filter outside the linguisitc process.

John noted that in some case if filter has no linguistic process we may end up with segments across trans-unit (for example when there are length limitation). Andrzej mentionned the case of 'merged attribute' and where multiple translation segments are needed. John underlined that sometimes the 'segment' has no linguistics aspects at all. Those maybe need to be hanled case by case while we can have more generic approach for other cases.
Andrzej mentionned his experience for this: non-equiv attribute (translation not equivalent to the source 'segment') and the need for the 'merge attribute'. Magnus said those mechanisms looked useful.
John asked if we had a case that was considering re-segmenting of a file. Magnus pointed out that the case 'Existing XLIFF files based on linguitic aspects' was probably touching on this aspect.

the group talked about the case where source/target are segmented and the translation tool need to re-segment them. Yves said that it was probably ok, as long as it happened within the same trans-unit. John and Andrzej noted that there would be possible issues with attributes at that level (e.g. one segment translated and another not).
Magnus said it would be useful to try looking at those use case with the different proposed representations. All agreed.


4.d) Discuss Implementation Options

Magnus listed the few ideas that were posted:

- [group and trans-unit] using <group> and multiple trans-unit (Tony)
- [segment element] using a new <segment> element within each trans-unit (Magnus)
- [tm namespace] using inline tags of another namespace, e.g. the tm namespace (Andrzej)
- [mrk] using <mrk> within each trans-unit (Yves)
- [merged attribute] John added that we should look at the 'merged attribute' case.

Magnus decided we need to look at the pros and cons in each cases. That was carried as actions items.


5. Any Other Business

None.

Next meeting in two week from now.



Action Items:

	* Andrzej: To investigate further issues related to using SRX with XLIFF.
	* Magnus: To look at pros/cons for send pro-cons for [mrk], [group and trans-unit], and [segment element].
	* Yves: To look at pros/cons for [tm namespace]
	* Andrzej: To look at pros/cons for [merged attribute]