I don't remember if I brought this up in the past, but some people have suggested to add a tool to our translation infrastructure for doing translations online. Some translators might prefer it, so we could have that too (as long as it integrates with our existing workflow).
Benefits: more people working on the same translation, lower barrier to entry, easier to handle big files, tools have workflow embedded (reviews etc).
To the folks that are interested in working on our l10n toolchain, here's a small task: Interested in investigating this a bit more? :)
Some Qs that we'd need answered: What tools are out there (eg. Pootle)? How do they interface with the VCS and what are the options for authentication? What are their requirements in terms of software and infrastructure?
-d
la, 2007-09-29 kello 14:33 +0100, Dimitris Glezos kirjoitti:
I don't remember if I brought this up in the past, but some people have suggested to add a tool to our translation infrastructure for doing translations online. Some translators might prefer it, so we could have that too (as long as it integrates with our existing workflow).
Benefits: more people working on the same translation, lower barrier to entry, easier to handle big files, tools have workflow embedded (reviews etc).
Lowering the barrier for entry is not necessary desirable. In my opinion a translator should understand what (s)he is translating, otherwise the translation may end up being words after another without a meaning.
Most Fedora utilities are quite technical. If a translator's knowledge is limited to using a web browser and writing e-mail, then things such as "iSCSI initiator", "volume group", or "BSD disk label" in Anaconda are likely to be quite remote and abstract to him/her. I am not saying the translator should be an expert on technicalities, but at least they should know the concept of partitioning a disk, and have some idea on how networks work, and so on.
Also if the barrier is too low, people with a lot of enthusiasm but not that much patience may turn up to translate a few dozen strings, and then disappear. To some it may be a big surprise that translating a whole application is real work, which requires time and thinking, and is not simply a matter of replacing English words with ones of one's own language.
-LN
Hello Lauri,
Lauri Nurmi wrote:
la, 2007-09-29 kello 14:33 +0100, Dimitris Glezos kirjoitti:
I don't remember if I brought this up in the past, but some people have suggested to add a tool to our translation infrastructure for doing translations online. Some translators might prefer it, so we could have that too (as long as it integrates with our existing workflow).
Benefits: more people working on the same translation, lower barrier to entry, easier to handle big files, tools have workflow embedded (reviews etc).
Lowering the barrier for entry is not necessary desirable. In my opinion a translator should understand what (s)he is translating, otherwise the translation may end up being words after another without a meaning.
I don't agree. I would like to translate through the web interface and I'm a technical person.
Most Fedora utilities are quite technical. If a translator's knowledge is limited to using a web browser and writing e-mail, then things such as "iSCSI initiator", "volume group", or "BSD disk label" in Anaconda are likely to be quite remote and abstract to him/her. I am not saying the translator should be an expert on technicalities, but at least they should know the concept of partitioning a disk, and have some idea on how networks work, and so on.
So, does ability to modify and commit a PO file results in fact that the person knows the concept of partitioning a disk? I don't think so, person can only follow translation guide to do so.
Also, my idea of this is that not everyone would be able to directly commit. In my head, everybody would be able to commit translation into "testing/rawhide" PO, after what the maintainer of the language would have to review the translation and approve it if it's correct.
In other words, less technical person would be able to help with translation and technical maintainer of the language will only have to review the translation and not translate from nothing.
Also if the barrier is too low, people with a lot of enthusiasm but not that much patience may turn up to translate a few dozen strings, and then disappear. To some it may be a big surprise that translating a whole application is real work, which requires time and thinking, and is not simply a matter of replacing English words with ones of one's own language.
It's why we have team leads for languages, which can take care and correct issues with translations. From my point of view, few dozen strings is better that nothing.
-LN
Just my $0.2 :)
Hello Dimitris,
Dimitris Glezos wrote:
I don't remember if I brought this up in the past, but some people have suggested to add a tool to our translation infrastructure for doing translations online. Some translators might prefer it, so we could have that too (as long as it integrates with our existing workflow).
Benefits: more people working on the same translation, lower barrier to entry, easier to handle big files, tools have workflow embedded (reviews etc).
To the folks that are interested in working on our l10n toolchain, here's a small task: Interested in investigating this a bit more? :)
+1, I think that this is what Fedora Translation project really needs. If I look at ubuntu (lunchpad) or jamendo.com, I see that this strategy works in awesome way!
Please, count me in for help!
Some Qs that we'd need answered: What tools are out there (eg. Pootle)? How do they interface with the VCS and what are the options for authentication? What are their requirements in terms of software and infrastructure?
- From system point of view, I'm not sure what's the best. But form I've heard pootle isn't so bad.
- Regarding authentication processes, I think everybody should be free to translate without login, but for direct commit only few people (team leads) would have ability to commit translated text.
- I would integrate it to transfix, what do you think? How much work is it?
-d
Thank you,
Στις 01-10-2007, ημέρα Δευ, και ώρα 16:17 +0200, ο/η Marek Mahut έγραψε:
Hello Dimitris,
Dimitris Glezos wrote:
I don't remember if I brought this up in the past, but some people have suggested to add a tool to our translation infrastructure for doing translations online. Some translators might prefer it, so we could have that too (as long as it integrates with our existing workflow).
Benefits: more people working on the same translation, lower barrier to entry, easier to handle big files, tools have workflow embedded (reviews etc).
To the folks that are interested in working on our l10n toolchain, here's a small task: Interested in investigating this a bit more? :)
+1, I think that this is what Fedora Translation project really needs. If I look at ubuntu (lunchpad) or jamendo.com, I see that this strategy works in awesome way!
Please, count me in for help!
Awesome. Since we've now got some experience in setting up similar software, I could put up a task page with some steps we need to take, and then we can see from there what we need.
Marek, are you willing to lead/push this? Of course I'll help any way I can and co-maintain.
Some Qs that we'd need answered: What tools are out there (eg. Pootle)? How do they interface with the VCS and what are the options for authentication? What are their requirements in terms of software and infrastructure?
- From system point of view, I'm not sure what's the best. But form I've
heard pootle isn't so bad.
- Regarding authentication processes, I think everybody should be free
to translate without login, but for direct commit only few people (team leads) would have ability to commit translated text.
From what I know, we've got this obligation that all translations should
be done by people who have signed the CLA.
So, we should check whether giving language coordinators the ability to commit non-CLA-signed stuff is a good idea or not. Theoretically, if the person is willing to take the responsibility, it should be OK. And I don't think we'll have a problem with translations (it's not like it's code or art).
Karsten? (CCing)
- I would integrate it to transfix, what do you think? How much work is it?
*Transifex*. :D
It will require a considerable amount of work. But the low-level parts are there: commit/push, integration with the Account System, controller examples. It's a really nice project for anyone who'd like to dig Turbogears a bit. Also, the foundation of Pootle (translate-toolkit) could be used here as well, and a project like this could give us a starting point for a workflow layer (translators, reviewers, etc).
Installing and extending Pootle (upstream, upstream, upstream!) seems the Right Thing for now; some research should be made to see if implementing "from scratch" would be more beneficial long-term.
-d
Dimitris Glezos wrote:
Στις 01-10-2007, ημέρα Δευ, και ώρα 16:17 +0200, ο/η Marek Mahut έγραψε:
Hello Dimitris,
Dimitris Glezos wrote:
I don't remember if I brought this up in the past, but some people have suggested to add a tool to our translation infrastructure for doing translations online. Some translators might prefer it, so we could have that too (as long as it integrates with our existing workflow).
Benefits: more people working on the same translation, lower barrier to entry, easier to handle big files, tools have workflow embedded (reviews etc).
To the folks that are interested in working on our l10n toolchain, here's a small task: Interested in investigating this a bit more? :)
+1, I think that this is what Fedora Translation project really needs. If I look at ubuntu (lunchpad) or jamendo.com, I see that this strategy works in awesome way!
Please, count me in for help!
Awesome. Since we've now got some experience in setting up similar software, I could put up a task page with some steps we need to take, and then we can see from there what we need.
Marek, are you willing to lead/push this? Of course I'll help any way I can and co-maintain.
Unfortunately no :( I don't have enough knowledge of python and turbo gears at all. I'm volunteering to help with other issues but I would like to leave python to experts. :)
Some Qs that we'd need answered: What tools are out there (eg. Pootle)? How do they interface with the VCS and what are the options for authentication? What are their requirements in terms of software and infrastructure?
- From system point of view, I'm not sure what's the best. But form I've
heard pootle isn't so bad.
- Regarding authentication processes, I think everybody should be free
to translate without login, but for direct commit only few people (team leads) would have ability to commit translated text.
From what I know, we've got this obligation that all translations should
be done by people who have signed the CLA.
So, we should check whether giving language coordinators the ability to commit non-CLA-signed stuff is a good idea or not. Theoretically, if the person is willing to take the responsibility, it should be OK. And I don't think we'll have a problem with translations (it's not like it's code or art).
I agree too, translation isn't personal propriety.
Karsten? (CCing)
- I would integrate it to transfix, what do you think? How much work is it?
*Transifex*. :D
Houps! ;)
It will require a considerable amount of work. But the low-level parts are there: commit/push, integration with the Account System, controller examples. It's a really nice project for anyone who'd like to dig Turbogears a bit. Also, the foundation of Pootle (translate-toolkit) could be used here as well, and a project like this could give us a starting point for a workflow layer (translators, reviewers, etc).
Installing and extending Pootle (upstream, upstream, upstream!) seems the Right Thing for now; some research should be made to see if implementing "from scratch" would be more beneficial long-term.
I will have a deeper look at pootle.
Anyone else interested to help with this project?
-d
Hi Dimitris, Marek, others
On Sat, 2007-10-06 at 00:55 +0100, Dimitris Glezos wrote:
Στις 01-10-2007, ημέρα Δευ, και ώρα 16:17 +0200, ο/η Marek Mahut έγραψε:
<snip>
It will require a considerable amount of work. But the low-level parts are there: commit/push, integration with the Account System, controller examples. It's a really nice project for anyone who'd like to dig Turbogears a bit. Also, the foundation of Pootle (translate-toolkit) could be used here as well, and a project like this could give us a starting point for a workflow layer (translators, reviewers, etc).
Installing and extending Pootle (upstream, upstream, upstream!) seems the Right Thing for now; some research should be made to see if implementing "from scratch" would be more beneficial long-term.
These comments are a relief, I really don't want to see l10n technical skills burned building another Pootle & Translate Toolkit if we can avoid it.
Just some quick notes of activity we're working on: * Optimisation work on the Translate Toolkit * Pulling much functionality out of Pootle into the Toolkit (this is essentially to allow us to move off of the currently unmaintained jToolkit web framework onto something else).
There are a number of things we are looking into going forward including workflow management.
Can I suggest perhaps a get together to exchange ideas with the Pootle developers. Lets rather work together to create good l10n infrastructure that isn't proprietary, doesn't burn our limited programming resources, builds on what we have and helps build tools that lower the barrier to entry for localisers.
-
Dimitris Glezos -
dwayne＠translate.org.za -
Lauri Nurmi -
Marek Mahut