Hi All,
Just back from vacation and catching up on email. Reading the current spec I think it could make sense to simply rely on section “2.6.8 Content Comparison”
in the core specification for what (if any) normalization to apply for validation comparisons. It stipulates that NFC is used to compare the equality of content.
Best regards,
Fredrik Estreen
From: [mailto:]
On Behalf Of Dr. David Filip
Sent: den 2 april 2013 14:49
To: Helena S Chapman
Cc: Kevin O'Donnell; Schurig, Joachim; Ryan King; ; Yves Savourel
Subject: Re: [xliff] R37: Revised Validations Module proposal
IMHO validation is different use case to size restriction. It makes sense to have none as default there but not here..
See inline..
Dr. David Filip
=======================
LRC | CNGL | LT-Web | CSIS
University of Limerick, Ireland
telephone: +353-6120-2781
cellphone: +353-86-0222-158
facsimile: +353-6120-2734
mailto:
On Wed, Mar 27, 2013 at 6:02 PM, Helena S Chapman <> wrote:
Unfortunately, I don't entirely agree.
I understand what you are suggesting with "language neutral rules". However, Unicode normalization should often be applied even when there is no language rules that should be associated with it.
+1
For instance, when someone send an English document which contains two distinctly different characters
"Ω" vs "Ω", without normalization and the correct intention, how would the tools/processes know what to do with these two characters? When to
treat them as two different characters and when to treat them as the same? "none" for normalization would tell them to treat these as different and if it's "NFD" or "NFC", these two would be the same (not identical).
This is going to be increasingly common because:
+1
1. We are more likely to receive true multilingual static content these days. And, we don't have to look far for an example. In Canada, most content has to be available both in English and French at the same time.
2. When we deal with multimedia type content, the use of more than one language within the same context is even more frequent. In my own household, a combination of Mandarin, Taiwanese, Japanese, and Hebrew are
often mixed in with English.
I am actually curious if the spoken language content interchange is out of scope of XLIFF in general? What happens when we embed this into an interactive format? Do we give our community the guideline that if one
is working with translation requests that are not limited to written languages, don't use XLIFF for interchange?
To include voice content we would need to re-charter. I think it is the next frontier and worth discussion.. I would just see it not at the front burner right now with 2.0 preparing for the first public review..
From: "Kevin O'Donnell" <>
To: "Dr. David Filip" <>,
Helena S Chapman/San Jose/IBM@IBMUS
Cc: "Schurig, Joachim" <>,
Ryan King <>, "" <>,
Yves Savourel <>
Date: 03/21/2013 04:30 PM
Subject: RE: [xliff] R37: Revised Validations Module proposal
Sent by: <>
Hi David,
This is a fair point. I’d be interested in hearing from tool providers about their preference here also.
I see two scenarios that are relevant here:
1. Global (language neutral) rules: these rules are the most common here at Microsoft and do not differentiate per-language (e.g. formatting rules). Therefore, rules of this nature
would not require/benefit from normalization
2. Language-specific rules: these rules, by default, may benefit from normalization and indeed would help reduce false positives if present
In my thinking, keeping “none” as default keeps the module simple and avoids unnecessary overhead by the processing agent when not required. If the XLIFF creator is implementing language-specific
rules, then they have an onus to specify their preferred normalization approach.
If we can surmise the likely prevalence of scenario 1 vs. scenario 2, that may also indicate the likely best default setting here.
Other feedback/thoughts appreciated.
Thanks,
Kevin.
From:
[mailto:]
On Behalf Of Dr. David Filip
Sent: Thursday, March 21, 2013 12:08 PM
To: Helena S Chapman
Cc: Schurig, Joachim; Ryan King;
; Yves Savourel
Subject: Re: [xliff] R37: Revised Validations Module proposal
Guys, after a brief spotcheck I just forwarded Asanka's minutes from Tuesday to the list.
While there was no formal conclusion the discussion tended to inclusion of normalization types along the lines of the size restriction module.
I think there was no doubt that we should provide a vehicle for conveying the normalization type required.
There did not seem to be a clear consensus on the default value though. I personally think that the default should NOT be "none". This option for default seems vague and obscure to me.. It lets the processor guess based on tribal knowledge what they should
do not to produce tons of false positives. I thought that we wanted to be naive implementer friendly.. :-)
Cheers
dF
Dr. David Filip
=======================
LRC | CNGL | LT-Web | CSIS
University of Limerick, Ireland
telephone: +353-6120-2781
cellphone: +353-86-0222-158
facsimile: +353-6120-2734
mailto:
On Thu, Mar 21, 2013 at 6:44 PM, Helena S Chapman <> wrote:
You summarized correctly of my own recollection of the discussion.
From: "Schurig, Joachim" <>
To: Ryan King <>,
Helena S Chapman/San Jose/IBM@IBMUS
Cc: ""
<>, Yves Savourel <>
Date: 03/20/2013 07:29 PM
Subject: RE: [xliff] R37: Revised Validations Module proposal
Hi Ryan,
while yours was my initial position as well, I do not think that it was the outcome of the discussion in the TC. We do have already mention of the normalization approach in the size restriction module, so it would make sense to include it here, too, and I think
this was the conclusion on the Tuesday call. You could leave the default to “none” and declare that this would leave it to the processing agent how to deal with the situation, but if any of “nfd” or “nfc” values are set it should lead to more specific behavior.
Could this be an acceptable solution to all parties?
Cheers,
Joachim
From:
[mailto:]
On Behalf Of Ryan King
Sent: Mittwoch, 20. März 2013 17:58
To: Helena S Chapman
Cc: ; Yves Savourel
Subject: RE: [xliff] R37: Revised Validations Module proposal
Yes, Helena, thanks for checking with me that. We did discuss it and feel that the processing agent should be responsible for normalization of text and so we will explicitly state that in the module.
Thanks,
Ryan
Sent from my Windows Phone
From: Helena S Chapman
Sent: 3/20/2013 4:43 AM
To: Ryan King
Cc: ; Yves Savourel
Subject: RE: [xliff] R37: Revised Validations Module proposal
Did Kevin convey the comments about normalization to you? How do we expect to deal with that in the spec?
From: Ryan King <>
To: Yves Savourel <>,
"" <>
Date: 03/20/2013 02:41 AM
Subject: RE: [xliff] R37: Revised Validations Module proposal
Sent by: <>
Hi Yves, all,
We suggest that for mustLoc, we enclose the source and replacement target values in parenthesis, like so: mustLoc="(World) (Welt)"
If for any reason, a parenthesis is required to be translated, as a brace for example, we could escape it like so: mustLoc="(\(World\)) ({Welt})"
Since we are generalizing the dblSpace to occurrences, then we could do something similar there as well: occurrences="(|) (3)"
For example, where 3 pipes need to occur in the target for whatever reason.
Further comments or suggestions welcome.
-----Original Message-----
From: Ryan King
Sent: Tuesday, March 19, 2013 10:59 PM
To: 'Yves Savourel';
Subject: RE: [xliff] R37: Revised Validations Module proposal
Thanks Yves for the feedback. All valid and good comments, which we will incorporate into the spec. As for the question on the mustLoc separator, Kevin and I are discussing it and suggest something shortly.
Thanks,
ryan
-----Original Message-----
From: [mailto:]
On Behalf Of Yves Savourel
Sent: Tuesday, March 19, 2013 5:01 AM
To:
Subject: RE: [xliff] R37: Revised Validations Module proposal
Hi Ryan, all,
Some feedback on the Validation proposal (nothing major, just possible suggestions):
-- strbegins and strEnds attributes:
Maybe names such as startsWith and endsWith may be a bit more descriptive of the function?
-- dblSpace:
This seems to be a very specific check. Maybe it can be generalized a bit without making it very different? For example, instead of dblSpace="3" we could do occurrence=" |3" (or a better name than 'occurrence'). This would allow to check for more than double
spaces.
-- mustLoc:
How do you represent the '|' if it needs to be in the left part of the value?
-- existsInSource, disabled:
So far I think XLIFF is using yes|no for Boolean rather than true|false. Maybe we could be consistent?
cheers,
-yves
---------------------------------------------------------------------
To unsubscribe from this mail list, you must leave the OASIS TC that generates this mail. Follow this link to all your TCs in OASIS at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
