Hi all.
The Docs Project is thinking of substituting entities like "Fedora 7" in all documentation with "&FC;", so that these entities won't change with each upgrade. For example, now we have in the PO:
This paragraph is about Fedora.
The change will make this:
This paragraph is about &FED;.
Is there any language that would have a problem with this? One such problem would be languages which change nouns ("Fedora") in different types of sentences, for example:
* "This is Fedora" * "These features of Fedora" * "Burn Fedora on a CD"
In greek we've left these untranslated and we won't have a problem. What about other languages? Is everyone OK with this?
-d
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Dimitris Glezos napisał(a):
Is there any language that would have a problem with this? One such problem would be languages which change nouns ("Fedora") in different types of sentences, for example:
- "This is Fedora"
- "These features of Fedora"
- "Burn Fedora on a CD"
In Polish it should be:
* "To jest Fedora" * "Te funkcje Fedory" * "Nagraj Fedorę na CD"
In greek we've left these untranslated and we won't have a problem. What about other languages? Is everyone OK with this?
In Polish unchanged nouns look very bad and simply ugly. I don't think it's acceptable.
- -- Piotr "Raven" Drąg http://raven.pmail.pl/
On Wed, 2007-04-04 at 22:45 +0200, Piotr 'Raven' Drąg wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Dimitris Glezos napisał(a):
Is there any language that would have a problem with this? One such problem would be languages which change nouns ("Fedora") in different types of sentences, for example:
- "This is Fedora"
- "These features of Fedora"
- "Burn Fedora on a CD"
In Polish it should be:
- "To jest Fedora"
- "Te funkcje Fedory"
- "Nagraj Fedorę na CD"
In greek we've left these untranslated and we won't have a problem. What about other languages? Is everyone OK with this?
In Polish unchanged nouns look very bad and simply ugly. I don't think it's acceptable.
Same problem with Serbian. It is not only bad, but also grammatically incorrect (leaving it untranslated might be ok for single word terms and trademarks). It gets worse for multi-word expressions like Fedora Documentation Project or Fedora Translation Project.
Cheers, Igor
So it sounds like we will have annoyances no matter which choice we make:
1. Using "xml2po -k" and keeping entities intact like &NAME; there are translators who will be inconvenienced. They can still substitute translations for the entity -- using the right form of "SomeName," instead of the exact string "&NAME;" -- but there will be no benefit to them in the event the entity changes. They may not even notice that it *has* changed unless they are watching the docs-common/common/entities/po/entities.pot file, and making the mental connection.
2. Using "xml2po -e" and transforming the entities into strings in the POT file, so "&NAME;" becomes "SomeName," it will be painful to change single strings in a minor way, since *many* fuzzy entries potentially result from trivial changes.
We can revert to behavior #2, and I can get this into the POTs for important modules now. I would like opinions within the next 24 hours. We first raised this issue almost two weeks ago in our meeting, summarized here:
https://www.redhat.com/archives/fedora-docs-list/2007-March/msg00186.html
Reminder to translators working on documentation: please make sure you are subscribed to fedora-docs-list@redhat.com for information. Thank you and we look forward to hearing from you about this issue.
O/H Paul W. Frields έγραψε:
So it sounds like we will have annoyances no matter which choice we make:
- Using "xml2po -k" and keeping entities intact like &NAME; there are
translators who will be inconvenienced. They can still substitute translations for the entity -- using the right form of "SomeName," instead of the exact string "&NAME;" -- but there will be no benefit to them in the event the entity changes. They may not even notice that it *has* changed unless they are watching the docs-common/common/entities/po/entities.pot file, and making the mental connection.
- Using "xml2po -e" and transforming the entities into strings in the
POT file, so "&NAME;" becomes "SomeName," it will be painful to change single strings in a minor way, since *many* fuzzy entries potentially result from trivial changes.
So, we (translators) need to decide here:
1. Following 1., some translators behave differently than the others and hard-translate the entities (substitute "&NAME;" with "Somename") when they need to. When an entity changes ("&FOO;" becomes "Fedora 8") the translator will have to manually update all occurrences for this language of "Fedora 8" to "Fedora 9".
2. All languages will get fuzzy translations once an entity changes. All of them will have to do the job described in 1.
So, the Q is: How often does an entity change and how big is this change? At this period we are changing "Fedora Core" -> "Fedora" and "Fedora Core 6" -> "Fedora 7". Later we'll probably only do the latter every 6 months.
If we choose 1, we can make sure that we notify those language teams that substitute the entities that the entities have changed. They can then search-and-replace all their files and substitute everything at the same time (with a tool like `regexxer`). So 1. might not be as bad as it sounds after all.
What do the translators of the languages in question think about this?
We can revert to behavior #2, and I can get this into the POTs for important modules now. I would like opinions within the next 24 hours. We first raised this issue almost two weeks ago in our meeting, summarized here:
https://www.redhat.com/archives/fedora-docs-list/2007-March/msg00186.html
Reminder to translators working on documentation: please make sure you are subscribed to fedora-docs-list@redhat.com for information. Thank you and we look forward to hearing from you about this issue.
Unfortunately I didn't catch this problem back then and I almost missed it some days ago because in greek we don't have it. More translators' eyes on -docs-list mean more chances to catch these problems on time. :-)
-d
On Fri, 2007-04-06 at 02:58 +0100, Dimitris Glezos wrote:
O/H Paul W. Frields έγραψε:
So it sounds like we will have annoyances no matter which choice we make:
- Using "xml2po -k" and keeping entities intact like &NAME; there are
translators who will be inconvenienced. They can still substitute translations for the entity -- using the right form of "SomeName," instead of the exact string "&NAME;" -- but there will be no benefit to them in the event the entity changes. They may not even notice that it *has* changed unless they are watching the docs-common/common/entities/po/entities.pot file, and making the mental connection.
- Using "xml2po -e" and transforming the entities into strings in the
POT file, so "&NAME;" becomes "SomeName," it will be painful to change single strings in a minor way, since *many* fuzzy entries potentially result from trivial changes.
So, we (translators) need to decide here:
- Following 1., some translators behave differently than the others and
hard-translate the entities (substitute "&NAME;" with "Somename") when they need to. When an entity changes ("&FOO;" becomes "Fedora 8") the translator will have to manually update all occurrences for this language of "Fedora 8" to "Fedora 9".
- All languages will get fuzzy translations once an entity changes. All of
them will have to do the job described in 1.
So, the Q is: How often does an entity change and how big is this change? At this period we are changing "Fedora Core" -> "Fedora" and "Fedora Core 6" -> "Fedora 7". Later we'll probably only do the latter every 6 months.
Some entities change less often, others more often. In the future I expect us to have a more robust test cycle for documentation, and you'll be looking at monthly changes for things like "test2" => "test3". A small change, but it would echo throughout many, many documents and message strings.
If we choose 1, we can make sure that we notify those language teams that substitute the entities that the entities have changed. They can then search-and-replace all their files and substitute everything at the same time (with a tool like `regexxer`). So 1. might not be as bad as it sounds after all.
-1 "we notify"; scaling problem +1 subscriptions to fedora-docs-commits@redhat.com
-
"Piotr \"Raven\" Drąg" -
Dimitris Glezos -
Igor Miletic -
Paul W. Frields