LISA: It's taken more time than originally expected to launch the much-awaited 1.4a release. As OSCAR's main TMX architect for the standard, can you tell us why it took so long?

Yves Savourel: Developing TMX is the collaborative work of a team: many have different opinions, sometimes opposed, and it takes time to get everyone to agree on things. But getting solutions everyone is comfortable with is important.

In addition, working during spring/summer time when many are enjoying some time off can also slow things down a bit. Overall the creation of TMX 1.4a took from mid-April to July-10 (about 3 months), not that long if compared to some other standard bodies. And in any case there is a lot to be said for Caesar's advice: "Festina Lente".

LISA: What are the main features of the 1.4a release?

Yves Savourel: The version 1.4a is essentially a reformulation of how inline codes should be dealt with in TMX. Except for the deprecation of , there are no syntactic changes. The 1.4a specification has some parts re-written and some added. The document itself has been re-formatted entirely. A new document and a set of test cases is also now available to verify compliance.

LISA: How does it differ from the previous versions?

Yves Savourel: The specification document lists all the changes (http://www.lisa.org/tmx/tmx.htm#AppChanges). To summarize: the only syntactic change is that the element has been deprecated.

LISA: We understand that TMX Certification is incorporated into the TMX standard release strategy. Please elaborate on what TMX certification implies, and how developers may comply with the standard?

Yves Savourel: Obviously a standard - any standard - is only as good as its implementations.

TMX follows that rule as well. A compliance kit is incorporated with the new version. This should help developers to implement solid and interoperable TMX functionalities.

Tools vendors can develop import and export functions so their applications can read and write TMX documents. Those TMX files must be valid, that is: well-formed XML that can be validated against the TMX DTD. However some aspects of the implementation cannot be verified by the DTD (for example: what type of inline elements the document uses to enclose inline codes). One way to verify a tool does a good job is to provide test case and check that the model TMX documents [in the compliance kit] are the same as the ones generated by the tool.

LISA: Is there a TMX "self-certification" category and how does it work?

Yves Savourel: Any tool vendor can verify whether its tool works with TMX 1.4a by going through the compliance test cases offered in the TMX Compliance Kit.

After that, one can choose to be certified - in other words to get LISA's stamp on the box of the product. Going through certification means an independent organization runs the compliance tests on the given tool. Such an organization must be accredited by LISA.

LISA: What are the different TMX certification levels?

Yves Savourel: TMX 1.4a has two levels: Level 1 is for TM with no inline codes (e.g., strings from a resource file), Level 2 is for formats that have inline codes (e.g., HTML content, where bold, italics, etc. are inline codes).

Depending on what type of original format you are working with, you should get TMX Level 1 or Level 2. A tool that offer HTML support but doesn't generate TMX document with inline codes is not TMX-compliant. Also keep in mind that tools may perhaps only import TMX or only export TMX (or do both). There are compliance tests for each of those aspects.

LISA: How will TMX certification benefit TMX users and clients?

Yves Savourel: Using certified tools should ensure that a user's TM assets will be more likely to be reusable across different tools. Because of the segmentation issue (different tools break segments differently), there is no way to be completely portable between tools, regardless of the TM format, but at least TMX will provide some interoperability.

LISA: Are their advantages for TMX developers to be certified?

Yves Savourel: Certified developers can make their customers happy by offering them a way in and out of their tools.

They can also offer a text-based format (vs. binary) to store TM databases in their tools - this will be handy for migration from one version to another. Supporting TMX also ensures that the users will be able to interoperate with other utilities if needed. Basically certification opens up tools.

LISA: How do the translation and localization industry benefit from standards like TMX?

Yves Savourel: In general standards such as TMX, OLIF, TBX, or XLIFF are good because they allow the users to have their assets - whatever they are - stored in a common and open format. This permits them to use various applications with the same data, and to migrate to newer and better tools without loosing too much data.

LISA: What can we look forward to see re/ the next version of TMX?

Yves Savourel: Actually I would hope we don't have too many new versions of TMX. The goal is to have a version that is as stable as possible, which has pretty much been the case since 1.3.

Some additional work to be done would be to provide an XML schema for TMX, in addition to the current DTD, so we can take advantage fully of XML features.

A possible addition, linked to XML Schema, would be to allow for non-TMX constructs inside a TMX document, using XML namespaces. This would be more flexible than the element and the ts attribute currently used for extensibility purpose.

And finally, there is the yet-to-be-resolved issue of segmentation. This is not a problem specific to TMX - it affects any TM repository and translation tool in general. Hopefully the Segmentation and Word Count Working Group newly created at OSCAR will be able to bring some solution to the problem, But this will take time.

Yves Savourel has been involved in internationalization and localization for more than a decade; first at International Language Engineering (ILE), then at Sykes Enterprises, and now at the RWS Group. His main focus has always been on developing tools and solutions for localization processes, often including SGML and XML aspects. Yves has been involved in the creation of various localization-related standards such as OpenTag, TMX, and XLIFF. He is also the author of XML Internationalization and Localization. He is a native of Brittany and lived in France, Africa and on islands of the Indian Ocean before settling in Boulder, Colorado.