I think discussions on the wall should be held in English. Only discussions in the comment section of a particular language might be held in that very language, since a reader who cannot read the comment might also not be interested in the sentence anyway.

pour l'instant la détection de la langue est faite via l'API google (cela pour des raisons historiques, nous attendons une refonte prochaine du site, pour passer sur nos propres algorithmes de détection automatique)
cela ne sollicite donc pas plus le serveur, mais préciser la langue rend l'ajout plus rapide vu qu'il n'y une chose en moins à faire. Après je n'ai jamais pris le temps de comparer les deux, pour voir si le gain de temps coté utilisateur était "observable".
l'option en fait est surtout pour les langues non supportés par l'API de google, ou produisant beaucoup de mauvaise détection (comme le shanghaïen, le latin etc.)

one more thing, we need to strictly define when does two sentences 'match', when is a direct link appropriate, when is an indirect link appropriate, and when an indirect link isn't acceptable.

There's way more linguistic nuances than just idioms and we really can't start being picky about every single one of them...

I don't know how 'official' my position is, but this is what I think.

Now, obviously, we'd like to have idioms included in the example sentences. We'd also like to have those examples translated. No controversy so far, I hope? ;-)

So, the point is that not every idiom will have an equivalent in every language it is translated to. Therefore it is impossible to always translate an idiom with an idiom.

Thus it must be acceptable, in some circumstances, to translate an idiom by a non-idiom phrase. Personally I see no reason why it should not always be allowed - but I prefer idom - idiom translations if they are available and natural.

> blay_paul pointed out what needs to be done to move the
> parenthesis to the right place

Except that for me it doesn't output like Paul pointed out... It outputs like this:
を <ruby><rb>監視</rb><rt>(</rt><rt>かんし</rt><rt>)</rt></ruby> し た 。

I have the same version of Firefox as he does. But I unfortunately don't have time to figure out what could be wrong...

As for your issue, I wasn't the one who implemented the MeCab part so I can't say if we can simply choose another atom. If not, we could tweak the output, but it's not something I can work on right now either.

the fact is the "educational" use exception imply that you attribute the source, it would not be problem if sentences where by tatoeba user on tatoeba for tatoeba only, the problem come the fact we release those sentences under a licence which authorize all possible use, which imply commercial and non educational use, moreover the export for the moment does not include attribution to the original author, so even for educational purpose, you need to attribute.

Big flaw of rating systems on Tatoeba? Too many sentences.

Personally, I prefer simple systems because they are less likely to break. I'm hardly the most active member of this community, but I fail to see the great threat from rogue contributors.

Should the community want to implement something like this, the four-step method you propose may work. However, apart from the obvious problem that you note regarding edited sentences there is also the problem that our content is different from Youtube in fundamental ways.

General users of Youtube watch particular videos and on topics that they generally have a very specific reaction to. The rating reflects that reaction; a very simple, natural measure to a particular experience.

General users of Tatoeba search for words or sentence patterns and will often browse through a number of sentences for what best fits their needs. I reckon that the greater number of items to rate will decrease dramatically the incentive to rate sentences (unless there is something especially wrong; in which case one would hope the user would be inclined to fix it anyway).

I'd rather see myself as the hard critic or devil's advocate than the pessimist, but I have my doubts that voting would be something the general user will do, leaving the task to the core community ... which is already working on this very issue by leaving comments (at least in the case of strange-sounding and unacceptable). What constitutes "very natural" sentences is going to differ between areas and might be better dealt with in comments.

On the other hand, should it be presented in a fairly seamless manner (e.g. with easy to use buttons on the search results page) and users react well to it, it could help users pick the best results. I'm furthermore not enough of a nay-sayer to stop well-meaning volunteers from working on areas that the majority is for.

That said, I'm still unconvinced about the need for this, reckon it will be too complicated if it is to be any good and fear it would be more of a distraction than of use. I think the current approach to adopt sentences and comment on that which seems odd is working well for the time being.

and too little 'active' contributors...and even less moderators...

I wonder if a simple flagging system based on tags would do...for example a button that's available to all (registered) users that when pressed adds a 'moderator attention' tag and needs a comment to be added first for the button to work...

Assuming the worst about human nature, I think the system would be more robust if it was a simple 0 to 100% rating, and if the "gravity" of a user's vote was a function of their contributions (quality and quantity) in that language. Youtube can never do something like this, but Tatoeba's structure would allow it.

This way, anyone could vote. For example, if twenty people all downrated a specific (say, French) sentence just to be nasty (let's say it was making light of the Bible, and they were all very pious Christians), but had no contributions in French, their votes would have 0 effect on the rating. This would put a natural feedback on bad voting behavior, and would automatically give weight to "good voters" (i.e. "good contributors").

Also, the rating system wouldn't be on the individual sentences, would it, but on pairs (i.e. the links between the sentences)? That would cover both the quality of translation and the naturalness...

Das ist leider nicht so einfach, sonst wäre es schon passiert.

It actually happened with Paul. He was away from Tatoeba and WWWJDIC for a whole year, and I wasn't sure if he'd ever come back. So what I did was transfer all this English-Japanese sentences to another user (fcbond).

There is no standard procedure to resign yet. There is also no way to check if a user is still interested in contributing... You have to resort to common sense.
If someone wants to stop working on Tatoeba, the best thing they can do is to inform the community about it, and to find other active members who are willing to adopt their sentences...

Anyway, as the community grows, it will definitely become more and more overwhelming for moderators, but we will adapt progressively. A few months ago, I was still the only one who could edit other people's sentences...
But in the long run, admins and moderators will not be the only ones with the permission to edit other people's sentences. Hopefully, we'll have a more advanced system where power and responsibilities will be better distributed, instead of hierarchical like it is now.

I agree that this has the potential to become a serious problem. However I think it can be alleviated significantly by some technical suggestions (some of which I've made before).

The comment system is no longer sufficient to keep up with requests for moderator action. The 'latest comments' disappear too fast, and it is too easy to miss ones that actually require moderator intervention - especially given the different languages involved.

I think there should be a specific way to mark comments as "action required", and all such comments should be included in a list until dealt with. The list should include sorting by date so that moderators will automatically know which ones have been waiting for two weeks or more without being dealt with.