[OSM-talk] Voting for adding the Translate extension to the OSM Wiki now open

Wed Apr 27 09:36:00 UTC 2022

Vào lúc 06:57 2022-04-26, Frederik Ramm đã viết:
> Hi,
> 
> On 22.04.22 23:55, Seth Deegan wrote:
>> Voting for the proposal to add the Translate extension to the OSM Wiki 
>> <https://wiki.openstreetmap.org/wiki/Proposed_features/Add_Translate_extension_to_Wiki> has now started.
> 
> I see the voting has prematurely been aborted due to new information 
> about the plugin.
> 
> My main concern for voting against this proposal was that, from what I 
> understood, the underlying idea is that
> 
> * all information about all tags should ideally be the same in all 
> languages
> 
> * if something diverges from "the global information" in one language 
> that diverging information must either be translated to English to 
> become part of the "the global information", or at the very least it 
> must be clearly identified which is the differing part so that it can 
> somehow be isolated in a language-specific subpage.
> 
> I am not sure if I have perhaps misunderstood something. One other voter 
> who was in favour of the plugin has written: "I am voting yes so that 
> the fledgling mapping community in Puerto Rico can more easily maintain 
> Spanish-language documentation that recognizes their country-specific 
> tagging nuances; so that developers in Vietnam don't need fluency in the 
> wiki's particular dialect of English in order to interpret our data 
> correctly; but most of all, so that we can try this out and see if it 
> works for us. Unlike with a tagging proposal, we will have a very 
> decisive solution if this turns out to be a terrible idea: uninstall the 
> extension."

I'm flattered that you found my comments important enough to selectively 
excerpt here. To be clear, I don't speak for the author of the proposal. 
In fact, someone who ultimately voted against the proposal had a bigger 
hand in shaping it than I did.

As I explained in a followup diary post [1], I'm primarily interested in 
solving existing shortcomings in the wiki's translation workflow and am 
not wedded to any particular solution. In my full vote comment and on 
the talk page, I made it clear that I had my own reservations about the 
proposal as written but came to this conclusion on balance. I recognize 
that others reasonably came to the opposite conclusion. Let's see if a 
revised proposal can find more common ground.

> The way I see it, the "fledgling mapping community in Puerto Rico" would 
> first have to analyse the differences between what their current 
> documentation says and what the English version says in order to then 
> produce a document that describes the differences. Where those 
> differences are nunances (for example, where the English version says 
> that something "is usually done" and the Spanish version says it "should 
> be done"), people will often not go through the effort to actually 
> retain these flavours because it would sound silly to write down 
> explicitly "for Puerto Rico, replace 'usually done' with 'should be 
> done'" or so).

I linked this phrase to a section of a tag description page [2] that I 
felt exemplified an opportunity for improvement. (In the future, I'll 
try to word my vote comments more carefully so that such links don't 
seem incidental when you quote me in plain text.)

In your vote comment, you gave the example of a Polish pedestrian 
crossing (which I assume is not a crossing whose nature depends on the 
language spoken by the pedestrian or mapper). There's no reason to stick 
to hypotheticals when there are plenty of concrete examples to consider. 
So I mentioned motorroad=*, which is tied to a specific road sign in 
specific legal systems, for which the wiki can only provide definitive 
guidance on a per-country basis. When this page was translated from 
English to Spanish in 2015, the source text was still a lot more sparse, 
and some countries' nuances were documented only in other languages. [4]

As things stand, someone updating the Spanish page could meticulously 
perform a three-way comparison between the original English, current 
English, and current Spanish to see whether there are any intentional 
differences. But more likely, they'd just start over from scratch, from 
English. By contrast, the Translate extension would avoid a loss of 
non-English information by showing the translator exactly what changed 
in the English version and what changed in the Spanish version of a 
particular passage, so they can respond to any differences on a 
case-by-case basis, however they see fit.

Puerto Rico literally shares the same road laws and road signs as the 
United States, but translated into Spanish. [3] In this case, it's a 
difference between "use this tag" in Spain and "do not use this tag" in 
Puerto Rico or anywhere else in the U.S. The translator wouldn't need to 
know how to drive in the U.S. to translate that section correctly.

While we're at it, did you know that the same kind of sign that Germany 
uses to legally reserve a road for cars is often used in Vietnam to 
legally ban cars from the road? Apparently the Vienna Convention is more 
malleable than one would suppose. I could've kept that tidbit hidden 
away in Vietnamese, but that would be preaching to the choir. It's 
everyone else who should be able to learn about that.

Obviously, not every tag is defined in terms of a legally binding road 
sign. But if the concern is that country-specific needs would be 
trampled upon, I'd contend that the Translate extension is not the 
villain, and you undermine your point by characterizing these real-world 
nuances as silly.

> In my mind, a move to this extension would massively impact the 
> understanding of the wiki; it would cement the dominance of English, and 
> every local information that diverges from the mainstream English would 
> have to be explicitly asserted as a deviation from the mainstream, 
> whereas currently someone can document something in their language 
> totally independent of what is said in other langauges.

To me, the fact that weight limits in Puerto Rico are signposted in U.S. 
tons in Spanish, or the fact that motorroads in Vietnam are reserved for 
motorcycles -- these are deviations from the mainstream, and no one 
should be ashamed of it. Quite the contrary: raising awareness of these 
cases lends them more legitimacy. How else would they be respected by 
tourist mappers, QA tools, and data consumers?

> This, in my eyes, represents a cultural change and the exact opposite of 
> what the voter above painted in rosy diversity colours.
> 
> Being a cultural, not a technical change, it will be impossible to 
> simply revert it by deinstalling the extension.

My point is that the technical changes described in the proposal are 
indeed reversible. The extension is essentially a fancy form that 
assembles the responses into a standard wiki page with invisible  comments scattered about. So if we were to uninstall the 
extension, any page that had adopted the extension would remain with 
exactly the same content. If uninstalling the extension doesn't remove 
the  comments automatically, a bot could perform a 
one-time find-and-replace across the wiki using a simple command. [5]

What isn't as easily reversible is that there would probably be more 
translations than before -- of more content in more languages. I expect 
less impact on existing translations than you do, in part because it's 
clear to everyone that some translators value the extra flexibility that 
comes from rolling a localized wiki page by hand. The proposal came with 
safeguards against overzealous language geeks, for what it's worth.

> (The same commenter said, in a different context, that a 50% majority 
> should be enough to install the plugin and anything else was setting the 
> hurdle too high, again claiming that you can simply uninstall the 
> extension if you don't like it. I think this is assumption is naive at 
> best - unless concrete plans are presented how all the changes made in 
> support of the extension would be rolled back with little effort.)

This is another mischaracterization of my comments. [6]

I think the tagging proposal process was ill-suited to evaluating this 
proposal. In the first place, it created a lot of confusion about 
whether tag description pages would be affected and how broadly the new 
system would be applied. The proposal contradicted itself on these points.

A reversible change of this nature should not require a 75% 
supermajority. After all, the threshold for *permanently changing the 
constitution* in many regions is a mere 60-66%. The tagging proposal 
threshold is so high because it's so hard to clean up an approved tag 
that later turns out to be a bad idea. Would we demand the same high bar 
for any technical change to the wiki, or just the ones that trouble 
certain people at the start of voting?

Regardless, I did not express a preference for a new threshold of 50%. 
It isn't my call. I just wanted to know if the proposer planned to stick 
to the 75% precedent they set, or if they would fall back to the 
"community consensus or at least a majority decision" that another 
administrator had originally asked for, or something in between.

> In order for me to support the installation of this extension, I would 
> have to have certainty that the aforementioned rosy language "the 
> mapping community in Puerto Rico can more easily maintain 
> Spanish-language documentation that recognizes their country-specific 
> tagging nuances" is actually true and not just an excellent exercise in 
> rhetoric. I want to see a concrete example how someone who speaks only 
> Spanish, only German, only Polish, can go to their language's 
> highway=path page and add some information there in their language...

I wasn't just spewing empty rhetoric. I'm actively documenting how to 
tag Puerto Rico traffic signs in both English and Spanish. [3] I've also 
contributed to documentation that started out in Vietnamese before being 
translated to English. All of this manually, because we're still using 
the same do-it-yourself approach to translations as in the early days.

Given this experience, the Translate extension seems tantalizing to me 
compared to the current situation, but show me a better solution and 
I'll support it. You also have some experience contributing to German 
documentation several years ago -- did that experience not strike you as 
even a little inconvenient? Would you allow those who wish to localize 
the wiki today to speak for themselves?

> ... WITHOUT having to navigate to a sub-section or sub-page that is 
> reserved for language-specific stuff ("you can't edit the stuff up here, 
> please go down there to add your information")
> 
> ... WITHOUT having to fear that something they added will be overridden 
> by conflicting information a second person has added to the English page 
> and a third person has translated to Spanish/German/Polish
> 
> ... WITHOUT having to first mentally dissect what the international 
> standard is and how whatever they want to add diverges or not from the 
> international standard and then clad it in appropriate markup
> 
> I'm willing to be convinced but frankly, what I've heard until now seems 
> largely marketing-speak and empty promises. "There will be some corner 
> of the wiki where people can document special cases" is not enough. In 
> fact, I even object to the idea that there is one standard documentation 
> and then possible "divergence" from that. One voter in favour of the 
> plugin writes "the LibreOffice wiki translation process is so much 
> better now" and this explains quite well what I feel - I think there's a 
> mindset that assumes OSM is some kind of software like LibreOffice and 
> everything is about describing the features of this software. But 
> tagging standards and conventions are the result of a social process, 
> not the description of the capabilities of some software.
> 
> And I think that everyone who says "let's just try that, we can always 
> go back" knows full well that it will only be weeks until it will be 
> impossible to go back.
> 
> If we want to install this for a "trial" then we should clearly limit 
> which pages it is to be used on (and those should not be tagging pages). 
> Use it for a description of the API or the data model, or the history of 
> OpenStreetMap or something. I can live with that.

If the proposer is open to this much less ambitious scope, then I'll 
take what I can get for now. :-) But it won't extinguish my hope that, 
someday, starting a localized wiki page of any kind won't be as daunting 
as it is today.

[1] https://www.openstreetmap.org/user/Minh%20Nguyen/diary/399068
[2] 
https://wiki.openstreetmap.org/wiki/Special:PermanentLink/2314672#United_States
[3] https://wiki.openstreetmap.org/wiki/ES:MUTCD/Puerto_Rico
[4] https://wiki.openstreetmap.org/wiki/Special:PermanentLink/1122173
[5] https://www.mediawiki.org/wiki/Manual:Pywikibot/replace.py
[6] 
https://wiki.openstreetmap.org/wiki/Special:PermanentLink/2315700#Comments_from_a_developer

-- 
minh at nguyen.cincinnati.oh.us