W3C home > Mailing lists > Public > public-multilingualweb-lt@w3.org > July 2012

Re: [all] readiness and translation process parameters

From: Felix Sasaki <fsasaki@w3.org>
Date: Thu, 5 Jul 2012 14:59:22 +0200
Message-ID: <CAL58czrxB6-GDZ+3ymL=wJwM70s9WfrkRSxAN3D+kJswi0VM6A@mail.gmail.com>
To: Arle Lommel <arle.lommel@dfki.de>
Cc: Dave Lewis <dave.lewis@cs.tcd.ie>, public-multilingualweb-lt@w3.org
2012/7/5 Arle Lommel <arle.lommel@dfki.de>

> Hi Dave,
> We have discussed specifying recommended values (via Linport), and we may
> yet do so, but there is no specific timeframe for this.
> What we keep running into (both in Linport and ISO) with this task, as
> with "domain" in our project, is that there is no single ontology for the
> permissible values. While some of the items (like source language) have
> obvious, enumeratable values (BCP47 in this case)

Actually this is not true: BCP 47 takes a generate approach for language
tags - there is a formal grammar for language tags
which can generate arbitrary language tags; only the subtags are taken from
this registry
which is also updated from time to time via proposals for new subtags. And,
there is an extension mechanism defined at

Sorry for nitpicking here, but you because of the above, you cannot
enumerate BCP47.

> , overall we do not believe we can enumerate the values.

I agree with this - and that's we reason I think this data category will
have no interoperable implementations. Dave, I can already tell you, with
this approach I would disagree with a "call for consensus". I understand
the wish to have this feature e.g. from Pedro, but my proposal would be

- Pedro's team  teams up VistaTEC or other LSPs
- Not our TC, but the LSPs tell us what their lowest bar for
interoperability is here.

> I'm not sure we can specify interoperability tests at this point.

Indeed - for me another reason to object against this moving forward.

> What we could do is specify that user agents present the values in an
> appropriate form. For machine-processible results we will have to wait (and
> I'm convinced that there will only be limited interoperability possible at
> best).
> I am not sure we will get the kind of LSP feedback we would like, but I
> can try to make some inquiries via Linport and GALA. It may take a while to
> get much of a response, however.

That's another reason to not to try to "push this feature into ITS 2.0".

> I've gone ahead and created an action for this:
> https://www.w3.org/International/multilingualweb/lt/track/actions/152
> I recognize that Felix won't want us to spend a lot of time on this for
> the moment, but I think this action will help us gauge interest in this
> topic in general and decide whether to invest more time for ITS 2.0 or to
> simply leave it for Linport. (I'm agnostic on this question at the moment.)

If Linport is already working on this, why doing it in ITS 2.0? The same
holds for specialRequirements IMO.



> Hope that helps,
> Arle
> On Jul 5, 2012, at 13:01 , Dave Lewis wrote:
>  Hi Arle,
> Thanks for that summary - the http://ttt.org/specs/ site is very helpful.
> However, you confirm my impression from this that OSI TS/11669 isn't
> defining permissable values for these, which limits the degree we can
> specify interoperability tests against these values.
> Are there plans for ISO to populate these values at some time, or are
> there any industry groups or plans at LINPORT to publish any best practice
> values for these attributes?
> Again it would be good to hear from the LSPs and clients on their interest
> in this spec so we can reach a decision on what level we can refer to these
> in either ITS2.0 or use them in test suites?
> cheers,
> dave
> On 04/07/2012 13:43, Arle Lommel wrote:
> Hi Dave,
>  The translation parameters are defined in ISO TS/11669, which was just
> published as an ISO technical specification. So those parameters are
> normative (at least if they apply to a task). Implementation is still a bit
> thin since the Technical Specification was only published a few weeks ago.
> However, they are the parameters that go into Linport, so there is
> increasing traction via Linport.
>  The fact that the parameter descriptions were published by ISO as part
> of TS/11669 (which has strict copyright restrictions and is sold), access
> would be a problem. But in this case Alan Melby negotiated an exception to
> general ISO copyright policy to allow the parameters *only* (not the rest
> of the Technical Specification) to be published openly. The URL is:
>  http://ttt.org/specs/
>  Right now these are not easily retrieved in a usable form (i.e., there
> should be a unique URI that retrieves the details for each parameter and
> nothing else, rather than the HTML page that is there now), but I can
> actually take care of that directly if I know what should be retrieved from
> a URI (e.g., the name, description, etc.) since I have access to the server
> to make these changes. Ideally, however, these would be referred to via
> ISOCat rather than from Alan's server, so I can see about how to add them
> to ISOCat (in which case the issue of access becomes part of the broader
> question of best practice for referring to ISOCat).
>  Note that you would still need name-value pairs at some level (although
> you might instead point to a URI with a full set of specifications
> [specifications = parameters + values]). So for the present, the reference
> to the URI would only give you the definition of the category, but not its
> value. So I would envision something more like this:
>   <its:rules
>   xmlns:its="http://www.w3.org/2005/11/its" <http://www.w3.org/2005/11/its>  version="2.0">
>  <its:transParam selector="/html/body" *transParamRef="http://www.ex.com/register.txt" transParamValue="formal"/>*
> </its:rules>
>  Not sure on that particular syntax, but it should get the idea across.
>  Note that ISO TS/11669 also does not define permissible values: Its
> categories were intended for human consumption, so if we want
> machine-processing interoperability, we would sill need a way to point to
> the permissible values in a given scenario. So the problem is only removed
> by a step, but we gain in this scenario in that *we* don't need to
> replicate anything in ISO TS/11669 and we open up whatever power is
> available from TS/11669.
>  Best,
>  -Arle
>  On Jul 4, 2012, at 14:08 , Dave Lewis wrote:
> Hi Arle,
> I agree, leaving this to Linport and ISO was the broad intent of that
> transParam proposal - they are the right people to do this.
> Further, I agree a dumb pointer to an external doc would be a reasonable
> alternative to name-value pairs - what do the LSP and client people think?
> Is this something people would be interested in implementing?
> As with the earlier transParam suggestion, this is soft on conformance,
> but a bit more specific in referring to Linport/ISO to solve this. But like
> the standoff provenance and PROV WG discussion, the maturity of this
> external work is a factor. Could you say a bit more about this normative
> status you mention? Can the documents be made available to the WG?
> I guess it would look like
> <its:rules
>   xmlns:its="http://www.w3.org/2005/11/its" <http://www.w3.org/2005/11/its>  version="2.0">
>  <its:transParam selector="/html/body" transParamRef="*http://www.ex.com/transParam.txt"*/>
> </its:rules>
> cheers,
> Dave

Felix Sasaki
DFKI / W3C Fellow
Received on Thursday, 5 July 2012 12:59:55 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:31:47 UTC