I see much better picture:
https://dl.dropboxusercontent.c...87/4355771.png
But, yes, the problem exists.

Well, maybe you're just too picky with fonts. :-)

This is how browsers on my Linux system render https://tatoeba.org/eng/sentences/show/4273602

Chrome
https://dl.dropboxusercontent.c..._chromiuim.png

Seamonkey (uses Gecko engine)
https://dl.dropboxusercontent.c..._seamonkey.png

Firefox
https://dl.dropboxusercontent.c...he_firefox.png

I don't see any noticeable difference.
Tatoeba displays sentences using "Trebuchet MS" or your preferred sans-serif font if the first one is unavailable on your system. Maybe you should check your font settings.

You're right, the problem affects not only the random sentence feature and my proposition is rather a workaround (and pretty harmless workaround if not useful, I believe) than a complete solution, but I think it would help to narrow the problem until it solved.

** Suggestion: Limiting the sentences displayed in "Random sentence" section to the languages defined in member's profile

It seems that a lot of new members are having problems understanding how the Tatoeba "Random sentence" feature works. If they see a sentence in a language they understand among the translations of the random sentence, they simply click "Translate" button without noticing that actually they translate the sentence on top of the list, not the one they maybe want to translate. The fact that all sentences but the one on top disappear after clicking the "Translate" button seems to never prevent many of new members from adding their translations. As a consequence we have a certain amount of wrongly linked sentences and most of such incorrect links won't be found and corrected soon enough.

I propose to limit the sentences that "Random sentence" feature displays to the languages defined in member's profile.

Well, if so, how about a corresponding option in user profile so everyone will be free to decide if he wants to see an additional button on not?

I think it's not necessary to make that extra button visible to everyone. As for corpus maintainers and admins, I'm sure they would be glad to have it.

I like the idea.

I've analysed Russian sentences with ZWS characters (38 in total) and found that all of them contains double ZWSes only. Most of those sentences was never edited. In the following example sentences ZWSes are represented by vertical line symbol (|):

#1052346 Прибыв на станцию||, она позвонила брату.
#1456323 Мать Тереза ||родилась в Югославии в 1910 году.
#1873268 Я уже помолвлена ||с моим другом, и я не нуждаюсь в верблюдах.
#1945261 Ты должен много заниматься, чтобы догнать свой ||класс.
#2134775 Он вложил весь свой ||капитал в бизнес.
#2281677 В Швеции есть свой ||собственный язык.
#2669291 Амазонка является второй по длине рекой в ||мире после Нила.
#2940499 С каждым днём ||рождения мы становимся старше на один год.
#3046001 Они отказались изменить свой ||образ жизни.
#3319065 Вы не могли бы дать мне свой ||номер телефона?
#3459211 Я ещё пока не принял решение||.
#3500791 Пятница - это день||, когда она очень занята.
#4155603 Почему ты не ешь свой ||шоколад?
#4171328 Хотите ещё ||стакан фруктового пунша?

In many sentences ZWSes appear in suspiciously similar positions, so there maybe some system. Also, 22 of 38 sentences belong to the same user.

There are some foreign proper names and loan words where "ё" is unstressed. For example, "сёрфингист", "сёгун", "Кёнигсберг".

Is it possible to make the search engine ignore combining acute accent in regular queries but take it into account if '=' sign before the word of interest was used?

Ah, now I see. Thanks, gillux!

That's strange. I haven't noticed any problem with alignment in my browsers while zooming in/out during tests. Can you show me how it looks like?

> I added a CSS trick to prevent the furigana to get into the copy and paste buffer, but it only works on Firefox.

Take a look at this dirty and ugly CSS trick. It seems to work on any modern browser. :-)
http://j-langtools.com/tatoeba/furiganatest.html

I'm not sure about Chinese sentences but as for the Japanese, it seems that the most of "その-sentences" come from Tanaka Corpus (most of English and Japanese sentences with numbers lower than ~30000 come from that corpus, actually). Tanaka Corpus is full of literal translations and less-than-natural sentences we often can see in textbooks where they are being used in educational purposes to illustrate different aspects of foreign languages. My Japanese is not good enough to judge if all those "その-sentences" are unnatural so let's wait for some native Japanese opinion. :-)

Russian plurals seem to work fine on http://dev.tatoeba.org.

Indeed, it seems that "other" was intended for decimals. For integer numbers new formula results only in 0, 1 and 2 just like the old one did.
There's no mistake in formula itself but I'm not sure if it will actually work for decimals (floating point numbers) since in most programming languages that distinguish number types "%" operator (modulus) takes integer operands.

The changes affect only Tatoeba user interface, not the sentences, so don't worry, you can continue using the Russian corpus as before. :-)

I've checked translations and corrected the problems you pointed out. Only phrases with {total} keywords had empty "many" cases. Phrases with {n} were unaffected.

Something is really wrong with plurals in Russian UI now. It worked like a charm before update but now it seems that plurals are being selected randomly and differ after every page reload.