mrk and overlapping cases

[Yves Savourel <ysavourel@enlaso.com>]

Hi all,

I've tried to implement <mrk> using the two notations for overlapping elements we discussed:

Overlapping example: <pc>aaa <mrk>bbb</pc> ccc</mrk>


Solution a) using two mrk linked by their IDs

<pc>aaa <mrk id='1'>bbb</mrk></pc><mrk id='1'> ccc</mrk>


Solution b) using the same start/end mechanism as in <sc>/<ec>

<pc>aaa <sm id='1'>bbb</pc> ccc<em rid=\"2\"/>


In my experience solution b) is a lot easier to implement. For example:

- I was able to reuse most of the code written for <pc>/<sc>/<ec>

- Using multiple <mrk> for the same annotation forces you to separate the markers from the data they refer to (since two spans can refer to the same data).

Obviously all this is highly dependent on the underlying implementation. Because I started mine with <pc>/<sc>/<ec> the data structure I'm using probably favor its re-use. So my test is probably bias a bit.

But even so, it looks to me that a notation consistent with <pc>/<sc>/<ec> has more advantages, or more exactly less drawbacks.

I'm also not working with a DOM object, but with the structure mapped to internal objects. That may change one's perspective too.

I have no idea how this would play with a completely different programming model, like using XSLT.
Bryan: any chance for you to give us some thoughts on what approach would be the easiest?
I'm asking you because I know you're using XSLT for your XLIFF tools, but maybe others have thoughts on that too? Or with other languages like Python?

Cheers,
-yves