Re: Updated Fresnel Ontologies and Examples - internationalization

From: David Huynh <dfhuynh_at_csail.mit.edu>
Date: Sat, 12 Mar 2005 12:48:16 -0500

Chris Bizer wrote:

>>Chris Bizer wrote:
>>
>>
>>
>>>Hi David,
>>>
>>>yes, good idea. What do you think about having a fresnel:contentNoValue
>>>property on the styleDescription class. Meaning you could say
>>>
>>>:knowsStyle fresnel:valueStyle [ fresnel:contentAfter ", " ;
>>> fresnel:contentLast ". "
>>>
>>>
>;
>
>
>>> fresnel:contentNoValue
>>>
>>>
>"This
>
>
>>>guy knows nobody." ] ;
>>>
>>>
>>>
>>>
>>It's hard for me to see how this fits in, as I'm not familiar with the
>>rest of Fresnel.
>>
>>On a different note, the N3 code you showed seems to make some
>>localization assumptions specific to English (or another similar
>>language). ", " assumes a left-to-right text flow and also assumes the
>>use of space (in Chinese, for example, there's no space between words).
>>In Chinese, I think you can also just juxtapose several nouns to make up
>>a list without having to use list separators.
>>
>>
>
>This is a hard one. No idea.
>
>:knowsStyle fresnel:valueStyle [ fresnel:contentAfter ", "_at_en ;
> fresnel:contentLast ".
>_at_en ] ;
>
>also doesn't help. Anybody else any solution?
>
>
As I said before, in Haystack, we have this duality of two display
primitives, "viewOne" and "viewMany". There are several implementations
of "viewMany" (e.g., as a list box, a graph layout, or a simple
sequential list). The simple sequential list "viewMany" knows how to
insert separators and terminator properly. It should be the thing that's
localized.

---
I think we need to push the "viewMany" stuff into lower layers of the 
system (whatever that means). Otherwise, here's an interesting bug that 
you will probably encounter (we did!): Say someone comes along and 
decide to put a little icon after each contact's name in a "knows" list:
    John Doe knows Alicia Hacker [icon], Ben Bitdiddle [icon], and 
Connie Cons [icon].
As the width of the line gets shorter, you might see the browser 
wrapping that line as follows:
    John Doe knows Alicia Hacker [icon], Ben Bitdiddle [icon]
    , and Connie Cons [icon].
or further:
    John Doe knows Alicia Hacker [icon], Ben Bitdiddle
    [icon], and Connie Cons [icon].
The right thing to do is to keep each [icon] on the same line as the 
last word of the corresponding name, and to keep the comma attached to 
the [icon]. But the browser can't do that unless it knows what each 
thing it's displaying is for and how they are related. This is sorta 
what we were trying to achieve with Haystack's UI framework--to let the 
semantics be connected from the datastore all the way to the rendered 
pixels, such that the browser is aware what each pixel represents.
(Yes, I realize I'm complaining about misplaced commas, but isn't RDF 
supposed to help us model tiny little details like this?)
---
Another note about lists: People might need a convenient way to display 
only the top X items and provide a link to the rest, or a way to show 
all items at once:
    A knows B, C, D, and [37 more people]
The truncation might not be specified as a number (3 as above), but 
might depend on the screenspace available (e.g., "show however many 
items that fits on at most 2 lines"). Yup, it's a tough one :-) but I'm 
very certain that it's needed.
---
Another feature request: It'd be nice if date values are automatically 
displayed in some more humane fashion than "Saturday, March 12, 2005 
12:35:09PM EST". "Today 12:35" would be better. Sometimes (and only 
sometimes), you want things like "5 minutes ago", "Last year", and "Last 
semester". Of course, in Haystack, we went all out and considered more 
semantic transformations that can occur before rendering, such as 
indicating an email my father sent me as coming from "My dad" rather 
than showing his full name. And if I send an email to both of my 
parents, the email would be shown as addressed to "My parents" than as 
addressed to two full names.
I believe that little details like these are really bringing home the 
point about RDF capable of modelling little things. And while you can 
already impress developers with huge chunks of RDF/XML, these little 
things are the things that impress the users :-)
David

Received on Sat Mar 12 2005 - 17:47:18 EST