On Tue, Jun 14, 2011 at 9:25 AM, Pierre-Yves Chibon <pingou(a)pingoured.fr&gt; wrote: Fedora could also host its own Transifex. The problem was to allocate resources to maintain this instance. That would be the same problem if we switch to Zanata. But as you said, it's only for test. -- Kévin Raymond User:shaiton GPG-Key: A5BCB3A2

On Tue, Jun 14, 2011 at 6:33 PM, Ruediger Landmann <r.landmann(a)redhat.com&gt; wrote: So let me take a different tact and ask: what problems are we trying to solve that Transifex doesn't or can't?

On Tue, Jun 14, 2011 at 12:33 PM, Dimitris Glezos <glezos(a)indifex.com&gt; wrote: I don't think so, this has not been discussed in the trans mailing list. -- Kévin Raymond User:shaiton GPG-Key: A5BCB3A2

On Tue, 14 Jun 2011 08:38:55 -0700 Toshio Kuratomi <a.badger(a)gmail.com&gt; wrote: I agree. Some additional thoughts: * We do not have any other jboss applications, so this would be all new to us, additionally, I don't think any of our main sysadmins have too much jboss experence. Maven, java and other parts of this are all things we do not currently run, so would be increasing the burden on all of us. * It's not clear to me that everything is packaged up in a way we require. I'm sure there are jboss rpms, but is everything else packaged? * I would really like to see a consensus from the fedora translation groups that this is something they would like to persue. We have had a long and happy cooperation with the transifex folks, and it would be good to discuss any issues with them to see if they can be solved. * asking translators to move to another completely new setup seems difficult when they just moved to transifex.net (all be it similar software). So, I really think the next steps here are to talk with transifex.net folks and the fedora translation community and see if there are issues that cannot be solved that this move would help with. kevin

On Tue, Jun 14, 2011 at 11:54:43AM -0400, David Nalley wrote: Insight is probably not a good example case -- it was hobbled by a number of roadbumps. A very dedicated and highly skilled team could do much better![1] Nevertheless I agree in general with David and Toshio. Is this RFR related to some larger roadmap for Fedora translation that's been proposed and reviewed by the Fedora translation community list? * * * [1] Which, as I understand it, is one of the reasons that we're currently using tx.net. There were very few people with skill and interest who stepped up to maintain internal translation services in Fedora, despite a clear ongoing need and requests for help. My recollection is most of those who did are either now working to make tx.net serve our needs, or are part of this conversation. -- Paul W. Frields http://paul.frields.org/ gpg fingerprint: 3DA6 A0AC 6D58 FEC4 0233 5906 ACDB C937 BD11 3717 http://redhat.com/ - - - - http://pfrields.fedorapeople.org/ The open source story continues to grow: http://opensource.com

On Wed, 2011-06-15 at 09:14 +1000, Ruediger Landmann wrote: We deflect all sorts of things precisely b/c "trying them out" is not a zero-cost operation. So - we're trying to figure out why we would commit resources to something if we don't have any thing to gain from it. -sv

On Tue, Jun 14, 2011 at 6:33 AM, Dimitris Glezos <glezos(a)indifex.com&gt; wrote: (Sorry for the late replies -- I'm still subscribed to this list from my personal account, and don't check it as often as I should.) I'm not aware of any specific concerns. If anything, I'm personally very happy with Transifex. That being said, since the transition from a self-hosted version of Transifex to an Indifex-hosted version of Transifex happened in the middle of the release cycle and without very much advanced notice, I agreed when we made the transition that we'd revisit the translation infrastructure decision after the Fedora 15 release, so I'm attempting to do that now. Red Hat's Engineering Services team in Australia, in the meantime, has obviously been working hard on Zanata (I think it used to be known by another name, which is why Rudi's email didn't catch my attention the first time). There may be compelling reasons to at least look at Zanata. There may no be. I don't pretend to know enough about Zanata yet to know what the real differences are, and I don't know any other way to find out other than to have a frank and open discussion with all the interested parties involved. So, let me lay my cards and the table and share what I do and don't know: 1) As I stated above, I'm personally happy with both the Transifex system itself and with the service and assistance rendered by Dimitris and Indifex. 2) Some people have expressed concerns to me personally that they're uncomfortable having a core piece of Fedora infrastructure hosted by a third-party, even if we have a great relationship with that third party. 3) At the same time, there's always been a lack of community involvement in the day-to-day maintenance (from a infrastructure standpoint) of Fedora's translation infrastructure. If we go back to any soft of self-hosted translation system (Transifex, Zanata, or anything else), it's going to need a small team of folks from the Fedora community to step up and dedicate themselves to keeping it running, keeping it patched, and keeping it upgraded. The ongoing maintenance burden is a something that keeps folks like me awake at night, and shouldn't be taken lightly. 4) Another thing to consider is whether a self-hosted solution has been packaged for Fedora. I know that Transifex is packaged for Fedora, but I don't see Zanata in the package list. 5) I know someone is going to suggest this, so let me squash the rumor before it has a chance to start. This is *not* a case of Red Hat mandating that we change systems. It's simply an effort on my part to do what we didn't have time for in the middle of the Fedora 15 cycle -- a chance to step back, look at the bigger picture of translation and translation infrastructure, look at the advantages and disadvantages of the various options, and make sure we choose a solid, sustainable set of tools. 6) While this concept is currently being discussed here on the infrastructure list, I think it's really a subject that needs to be discussed with a wider audience. It's a topic that affects infrastructure, translation, packagers, and documentation at a minimum. My plan is to kick of some discussions in those groups, and then try to schedule a wider meeting in the next two weeks (similar to the meetings we held before the transition to tx.net). -- Jared Smith Fedora Project Leader

On Thu, Jun 16, 2011 at 10:11:19AM -0400, Jared Smith wrote: The thing about this is that I don't think anyone on the present infrastructure team sees this as a problem; unless I've forgotten someone I think that we're unanimous in thinking of this as an opportunity. So the foundation of this argument carries less than zero weight with us (less as this argument just gets people to feeling that the people proposing the change are doing so because of fundamental disagreements that we'll never agree with.) Any change to production infrastructure will need to have reasons that transifex.net is a less desirable solution than something else that we manage. Noting here that you earn a greater degree of trust and freedom to make decisions when you join the Infrastructure team and happen to focus on a project rather than working on a project within infrastructure. The shift in emphasis there is whether you are around on IRC, contribute to other areas that aren't your primary focus, and generally, have a presence that says, I'm not just going to disappear and leave my projects resting on your shoulders in the future. Zanata is built on the JBoss framework and JBoss hasn't even been packaged in Fedora. So... having been a part of the multi-year insight development from the infrastructure side, there's two aspects: Consuming groups: 1) What features do we need? 2) What features do we desire? 3) What manpower can we pony up to deploy this? 4) What manpower can we pony up to maintain this? 5) What is out there that might satisfy our needs? Providing groups (infra): 1) What computing resources are needed to support this? 2) What manpower resources are needed to support this? 2.a) What security implications does running this entail? 2.b) Do we have inhouse knowledge that will support running this? 2.c) Do we have in-Fedora knowledge to fall back on? 2.e) Is this something we need to run? 3) What is the effect on Fedora if we find that we cannot continue to support this in the future? 4) What is out there that's going to be easiest to support? In the early stages, the consuming groups are really doing the most evaluation since they have to determine what features they must have and what features they desire to have. However, they either need to understand the support burden so they have something decent to present to infrastructure or (probably preferably) they need to stop their evaluation short of trying actual products and get someone in infrastructure interested in their list of features at that point so that they can evaluate the choices together. -Toshio

Thanks all for the feedback and comments; some of it has been extremely useful. I concur that it's premature to tie up Fedora resources in such a test; I'll investigate other hosting avenues and reopen the discussion here only if Zanata looks promising and after I have some buy-in from L10N. Cheers Rudi

On Tue, 21 Jun 2011, noriko wrote: They don't notice the discussion because no one has taken it there. I'm not sure why we're even still discussing this here since not a single email has been sent to http://lists.fedoraproject.org/pipermail/trans/ wrt Zanata. It's not free for us to just hand over resources that could be better planned. So I suggest again, start with the translators. If someone on the infrastructure team is willing to sponsor it I'm sure they'll speak up but it'd be a much better sell if the translation team actually got together and said "sure, lets take a look" -Mike

On Tue, Jun 21, 2011 at 11:19 AM, Kévin Raymond <shaiton(a)fedoraproject.org&gt; wrote: I also think we should have good documentation before requesting time to try out a tool -- eg. how does it compare with alternative tools such as Transifex, Pootle, Damned Lies, especially in the areas that are important to Fedora. -d -- Dimitris Glezos Transifex: The Multilingual Publishing Revolution http://www.transifex.net/ -- http://www.indifex.com/

On 06/22/2011 05:13 AM, noriko wrote: Yes - you essentially want Infra to agree to/"approve" something that we don't even know that people want to try out, because still to this day I do not see anything on trans@ even asking if they are remotely interested in this. My opinion here, but I think it would be wise to let this drop until that happens, and once it does, and a group from trans@ is actually interested and wanting to try this out, a new thread can open at that point. Right now, you're asking Infra to devote resources to a project that we don't even know that the core group using it will even want to try. There is a translation solution in place, and I've not seen anyone list problems with that one that Zanata solves. The point is, going back and forth on this without several things happening is pointless: 1) See if there is any interest in trans@ 2) If such a group forms, have them compile a list that clearly states what Zanata solves that TX (our current solution) does not. 3) At that point, it would probably be safe to open a new thread here, including that list, and seeing if it's worthwhile to pursue this. Until those things happen, this is seemingly just going to go back and forth without anything happening. Multiple people from infra have mentioned what needs to happen for this to go any further, and I see no one taking those steps. As Mike strongly hinted a few emails back, please let this one drop until that happens.