Re: Comments on "TML proposal"

Hi,
First Serge, thank you for raising this point.

I can confirm that we'd be very happy to work with you on a follow up 
article explaining how this functionality is supported in ITS2.0, with 
pointers to some of the implementations that are out there.

Also, this is not the end of the story, and the community have already 
been discussing issues like a common API for text analysis services and 
how user validation  of text annotation (positive and negative) can be 
fed back via such an API. It would be interesting to discuss the need 
for such further technical consensus building with the TAUS community.

kind regards,
Dave

On 08/04/2015 23:39, Serge Gladkoff wrote:
>
> Hi Brian, Jaap,
>
> Please, let me to step forward with a comment on the “TML proposal” 
> that has been published in recent article in TAUS Review #3 on 
> 'translation mark-up'  language:
>
> http://issuu.com/tausreview/docs/tausreview-dataissue-april2015/14
>
> The proposal suggests an annotation mark-up.
>
> On “volunteer mandate” from my fellow members of W3C ITS interest 
> group, please let me to point out the following:
>
> The TML seems to replicate ITS2.0 terminology and text analysis tag 
> functionality. ITS 2.0 is (just for the reference) as W3C standard 
> which has been developed recently by representative group of industry 
> experts (http://www.w3.org/TR/its20/).
>
> Your TML idea seems to be a simplified approach of similar nature.
>
> Well, ideas are hanging in the air, and Kudos to you for coming up 
> with the idea that has been good enough to develop the W3C standard, 
> recognized by well-deserved support of EU funding!
>
> However, ITS2.0 is well developed, stable standard with lots of 
> implementations and case studies already, not to mention the group 
> used that has developed it.
>
> AND, actually, looking at the proposal in more detail, my colleagues 
> have been quick to point out that “TML” as proposed has some real 
> problems in that it exposes metadata in ways that would totally break 
> in non-aware processes. Because it is relying on native HTML 
> mechanisms that only work if tools are of the “kludge” used to hide 
> the data, it would end up blowing up word counts for other tools, 
> result in content being accidentally translated (since <span> tags are 
> not, by default protected), etc. We really do not like the overloading 
> of existing mechanisms in this way when they serve fundamentally 
> different purposes. It would take only one tool in the chain that 
> processed the content in an intelligent way that does not match the 
> expectation for things to break.
>
> There is a real advantage to using ITS in these contexts because then 
> tools that don't understand it also know that they don't understand 
> it. The proposed TML in the article however, would be problematic 
> because there is nothing in it to tell the tools (a) that they don't 
> know what to do with it and (b) /what/ they should do with it. It is 
> thus entirely opaque in a bad way.
>
> ALSO, the „embedding comments“ section looks like a replication for 
> „Localisation note“. All three replications seem to be nearly exact 
> copies of ITS data categories. So for enabling ITS processors to 
> understand the proposed markup, one could create an ITS rules file. 
> You write „adding support will be a minimal effort“, but that has the 
> drawback that e.g. no validation is possible - which is easy with 
> validator.nu . Also, this only covers HTML5 - what about Dita, 
> docbook, mallard, TTML, XLIFF, ... (ITS spans across all these.)
>
> Finally, here’s FAQ on the HTML „translate“ attribute:
>
> http://www.w3.org/International/questions/qa-translate-flag.en
>
> Does TML have a translate marker? That would be an overlap with the 
> HTML native „translate“ attribute.
>
> Besides this maybe useful: this section
>
> http://www.w3.org/International/questions/qa-translate-flag.en#stickyness
>
> using class attributes, like TML does, to provide such metadata, is 
> described as legacy approaches. See esp.: "replacing 
> a class attribute value with an attribute that is a formal part of the 
> language makes this feature much more reliable, especially in wider 
> contexts.“
>
> Consequently, we would say that we would really recommend to use ITS 
> in these contexts. It is important that ITS effort is promoted, since 
> it serves to better purpose of interoperability and decrease of 
> duplication of effort in developing similar but incompatible solutions 
> in different corners of the industry.
>
> Jaap, is there a way to make sure that this feedback is propagated to 
> the TAUS Review readership?
>
> Perhaps we can arrange W3C ITS Interest Group to write a small 
> “counterpoint” article in the next issue of TAUS Review? Arle and Dave 
> Lewis could contribute with the language of the article.
>
> Very Best Regards,
>
> Serge Gladkoff
>