RE: [xliff-inline] Proposed requirement for inline SC: XML well-formed-ness as a design goal for XLIFF 2.0 inline markup

[Yves Savourel <ysavourel@translate.com>]

Hi Bryan,

>> frankly I don't know how we could represent '<' in
>> XML content as anything but some kind of
>> escaped sequence.
>
> I am not objecting to the practice of preserving code. 
> I am specifically objecting to the practice of escaping XML markup.
> Escaping XML markup (<bpt ctype="italic">&lt;i&gt;</bpt>yippy<ept>&lt;/i&gt;</ept>)
> is nothing but syntactic sugar. 
> ...

If '<' is part of the original code and if that original code is stored in a <bpt>, it must be escaped. It's not an optional choice. ('>' is optional). http://www.w3.org/TR/REC-xml/#syntax (short of using a CDATA section).



> Omitting the escaped XML (in this example) no less clear
> (<bpt ctype="italic" />yippy<ept />, of course in both examples
> the required id would need to be included).

I see now: what you seem to be alluding to is whether putting the original code is needed/wanted or not. A fair question, but (I think) un-related to escaping XML. That is, I don't think the decision of putting the native codes inside <bpt> should be driven by the type of native format or if it will need to be escaped. To me it's a decision based on how the tool deals with native codes: some for whatever reason need to preserve it, while other do not.

I think the bigger problem, beyond escaping things, is that using something like <bpt> means the text content of the segment is sometime real text, sometimes native code. That is if we were to do a getText() on the segment node the text return would be a mix of real text and native codes. Ideally we would want only real text.


> ...We use XML because XML is the mechanism we chose to solve 
> the I (interchange) challenge. If we identify escaping the 
> XML as a good practice, we're a bit hypocritical.

I'm afraid I'm still not grasping the escaping issue to its full measure yet :) For me it's just a side effect, a symptom. The real cause is that we have constructs like <bpt> instead of <g>. Maybe we will find a solution this time around...


Have a good evening Bryan,
-ys