Segmentation Issues in XLIFF

Hi Tony,

Thank you for your kind reply and the invitation to join the Technical Committee. I am very interested in contributing to this development, and TRADOS will be looking closer into our different options for joining OASIS.

I am glad to hear of your initiative to advance this topic. I have no objections to you forwarding my thoughts on this matter to other members of the committee or the public XLIFF mailing list, and I would be very interested in hearing opinions, input, and ideas from other members.

Regards,

Magnus

From: Tony Jewtushenko [mailto:Tony.Jewtushenko@oracle.com]
Sent: Thursday, January 29, 2004 11:44 AM
To: Magnus Martikainen
Cc: Yves Savourel; Jochen Hummel; John Reid
Subject: Re: Questions on XLIFF

Hi Magnus:

I know of you and Jochen by reputation, and it's a pleasure to make your acquaintance.

I agree with whole-heartedly with Yves that your observations and proposals on segmentation and XLIFF are very interesting and useful. To date the TC has very consciously set aside attempting to define a segmentation solution for XLIFF, but recent release of SRX and increased (TM) tool developer attention means that we must address this problem sooner rather than later..

XLIFF 1.1's XSD permits "namespace=##other" extension at the trans-unit level, so that means we could not extend using our own namespace. But we could extend with an external private namespace, which could be defined by committee so as to ensure cross-tool interchange of XLIFF data will be possible. To advance this issue formally, I suggest the XLIFF TC create a subcommittee whose goal it is to discuss, model and ultimately propose solution(s) for the segmentation problem. This will help move things along more quickly than attempting to work this through the open committee, since as Yves mentioned our resources are presently focused on developing XLIFF representation profiles for popular resource types.

Clearly your input (and by that I mean Trados' in general and yours in particular) to solving the segmentation problem would be valuable, so I invite you to join our Technical Committee. Membership is through the OASIS consortium, which provides a number of membership options. If you would prefer not to join OASIS, there may be other options available for us to work together, possibly as a joint committee with OSCAR SRX, but I'm not sure how quickly that could be arranged or structured. Past experience with joint-committtees makes me a bit of a pessimist in terms of expected productivity, so this would not be my preferred avenue. There may be other ways we can work informally as well.

With your permission I'd like to forward the earlier email discussion between ourselves to a subset of the TC committee, and ultimately to the public XLIFF mailing list. If you have any objections, could you voice them as soon as possible?

I look forward our working together to solve this problem.

Regards,
Tony

Yves Savourel wrote:

Hi Magnus,

Your ideas on the segmentation representation look very interesting. I agree

also on the different points you made. The two areas of potential problems

that I think will be the most difficult to deal with would be:

- Compatibility. Hopefully we should be able to come up with a mechanism

that will allow a seamless backward compatibility. Technically your example

could be already represented in XLIFF 1.1 using a different namespace for

the <segment-group> elements, and that namespace using XLIFF namespace

itself (I think, although someone with better grasp on the use of XSD <any

namespace=##other"> vs. <any namespace="##any"> should check that).

I'm not sure how current tools would deal with it though.

- As you also mentioned, the potential change of <ept> and <bpt> elements as

their content is broken down could also be a little tricky.

I'd like the input of the other XLIFF users/developers. As I'm often using

Trados tools, or Trados-like tools, my view of the translation process is

similar to yours. But other angles would be nice to have, often it allows to

come up with better ideas.

The TC is currently working on various "profiles" (how to represent various

formats into XLIFF), but segmentation will have to be addressed one way or

another at some point. The sooner we start exchanging ideas, maybe coming up

with a namespace for it, which later could be simple integrated in XLIFF's

core, or anything that can make things progress, the better.

Tony, John: any thoughts? Where should we go from here?

Cheers,

-yves

--

Tony Jewtushenko               mailto:tony.jewtushenko@oracle.com

Principal Product Manager      direct tel: +353.1.8039080

Oracle Corporation, Ireland

xliff message