MoinMoin: MoinMoinChat/Logs/moin-dev/2006-06-18

2006-06-18T02:04:45  <xorAxAx> ThomasWaldmann: what do you think about integrating an openid server/client into moinmoin?
2006-06-18T02:04:56  <xorAxAx> ThomasWaldmann: i somehow have to solve the authentication problem in my sync setup
2006-06-18T02:05:11  <xorAxAx> i.e. users want to sync with other wikis where the code needs to log them in
2006-06-18T13:02:40  <ThomasWaldmann> xorAxAx: as long as it doesn't make ext. dependencies or it's optional, it would be nice
2006-06-18T13:04:24  <ThomasWaldmann> but maybe first get sync working (with simple shared secret dict interwiki: secret), then do openid
2006-06-18T13:16:37  <xorAxAx> yes, good idea
2006-06-18T13:55:07  <ThomasWaldmann> Fabi: moin. did you already discuss implementation with xorAxAx?
2006-06-18T13:57:54  <xorAxAx> oh, he is online
2006-06-18T13:58:05  <xorAxAx> Fabi: yeah, have you looked at my wiki page?
2006-06-18T14:00:26  * ThomasWaldmann has to go. bbl.
2006-06-18T14:10:18  <Fabi> I saw you edited it, but did not yet read it in detail
2006-06-18T14:10:25  * Fabi starts reading
2006-06-18T14:11:02  <xorAxAx> if i would use mercurial, i would need to reinvent a few parts of the logic
2006-06-18T14:11:22  <xorAxAx> esp. because a wiki sync works page-wise and not changeset-with-multiple-files-wise
2006-06-18T14:11:43  <xorAxAx> this makes the amount of time i would need to spend quite much
2006-06-18T14:13:13  <Fabi> so you are planning to do a from scratch implementation?
2006-06-18T14:13:26  <xorAxAx> no
2006-06-18T14:13:41  <xorAxAx> i need to decide if i should choose my first or my second idea
2006-06-18T14:13:57  <xorAxAx> using that dvcs idea, i need to implement merging and pulling/pushing from scratch
2006-06-18T14:14:09  <xorAxAx> which would involve copying of code, though
2006-06-18T14:14:25  <xorAxAx> but i fear that i cannot estimate the needed amount of time and the obstacles
2006-06-18T14:14:48  <xorAxAx> compared to the first idea, which is rather easy but also limited in scalability
2006-06-18T14:15:08  <xorAxAx> do you understand the sketched problem of my first design?
2006-06-18T14:16:35  <Fabi> no 100% shure
2006-06-18T14:16:47  <xorAxAx> do you have a question? :)
2006-06-18T14:17:34  <Fabi> "How does synchronisation of 3 wikis work?"
2006-06-18T14:18:17  <xorAxAx> with the 1st or 2nd concept?
2006-06-18T14:18:19  <Fabi> and at which point do you need these (InterwikiMoniker, LocalRevision, RemoteRevision) tags
2006-06-18T14:18:30  <xorAxAx> ah, then its the first concept
2006-06-18T14:18:40  <xorAxAx> tags are important to mark merge points with other wikis
2006-06-18T14:18:40  <Fabi> are they stored in the RC of the pages
2006-06-18T14:18:48  <xorAxAx> they are stored parallely to the edit-log
2006-06-18T14:18:55  <xorAxAx> and are meta-data therefore
2006-06-18T14:19:16  <xorAxAx> they are used to determine the ancestor revision for a 3-way merge
2006-06-18T14:19:24  <xorAxAx> should be clear, right?
2006-06-18T14:19:30  <Fabi> yup
2006-06-18T14:19:41  <xorAxAx> ok, now the 3 wiki problem
2006-06-18T14:19:55  <Fabi> we need to ask the original wiki for this anchestor, right?
2006-06-18T14:20:09  <xorAxAx> yeah, that would solve it somehow
2006-06-18T14:20:12  <xorAxAx> but thats not implemented
2006-06-18T14:20:23  <xorAxAx> and it would make it much more complicated ...
2006-06-18T14:20:38  <xorAxAx> the current concept simply generates a conflict and lets the user clean up :)
2006-06-18T14:21:22  <Fabi> does it always generate a conflict?
2006-06-18T14:21:25  <xorAxAx> no
2006-06-18T14:21:37  <xorAxAx> conflict == 3-way merge didnt work cleanly
2006-06-18T14:21:45  <xorAxAx> a conflict is shown in the recentchanges
2006-06-18T14:21:55  <Fabi> yup, we can't solve that
2006-06-18T14:22:42  <xorAxAx> another question is: how often would one have a multi-wiki setup?
2006-06-18T14:22:45  <xorAxAx> with more than 2 wikis
2006-06-18T14:23:30  <xorAxAx> and not all scenarios are problematic ...
2006-06-18T14:24:26  <xorAxAx> so you think that the first concept is good enough for wikilandia? :)
2006-06-18T14:26:09  <Fabi> first concept works fine if all sync to a central instance, right?
2006-06-18T14:27:02  <xorAxAx> yep!
2006-06-18T14:27:29  <Fabi> but you get into trouble as soon as slave instances sync each other
2006-06-18T14:27:31  <xorAxAx> as i wrote, node reordering might result in conflicts
2006-06-18T14:27:39  <xorAxAx> yep
2006-06-18T14:28:03  <xorAxAx> (node = instance)
2006-06-18T14:29:42  <Fabi> I think is still did not yet understand what additional data we need to get from concept 1 to concept 2
2006-06-18T14:30:14  * Fabi looks for pen and paper
2006-06-18T14:33:02  <Fabi> hmm, things get complicated soon
2006-06-18T14:33:39  <xorAxAx> the DVCS model implemented by git/mercurial etc. is quite lightweight and scalable
2006-06-18T14:35:37  <Fabi> they add the history of the merged changes to the local history, right?
2006-06-18T14:35:56  <xorAxAx> merges are changesets, yes
2006-06-18T14:40:15  <Fabi> I think having a central master repository should be ok
2006-06-18T14:41:00  <Fabi> but we should make shure that only sane syncs can be done without force
2006-06-18T14:41:22  <xorAxAx> hmm
2006-06-18T14:42:00  <xorAxAx> ok, if page exists on both sides and tags are there, require the force flag :)
2006-06-18T14:42:25  <xorAxAx> tags are NOT there i mean
2006-06-18T14:42:35  <Fabi> (use the force, luke)
2006-06-18T14:43:11  <xorAxAx> yep
2006-06-18T14:43:16  <Fabi> I think about a triangle situation
2006-06-18T14:43:35  <Fabi> A<->B and A<->C
2006-06-18T14:43:43  <xorAxAx> yes
2006-06-18T14:43:59  <Fabi> so we need to avoid B<->C
2006-06-18T14:44:19  <xorAxAx> but the interesting thing is that B->C is allowed after the tags have been generated
2006-06-18T14:44:35  <xorAxAx> e.g. you just have to force the sync and then you can sync without force and problem
2006-06-18T14:44:39  <xorAxAx> s
2006-06-18T14:45:37  <Fabi> and syncing C back to A still works
2006-06-18T14:45:44  <Fabi> ?
2006-06-18T14:46:16  <xorAxAx> yes, why not?
2006-06-18T14:46:22  <Fabi> ok, fine
2006-06-18T14:46:31  <xorAxAx> at least i cannot think of a problem here
2006-06-18T14:47:04  <Fabi> I think this is good enough
2006-06-18T14:47:12  <xorAxAx> ok :)
2006-06-18T14:47:22  <Fabi> wikis are not the linux kernel where you need lots of branches
2006-06-18T14:47:32  <xorAxAx> right, many  branches are bad
2006-06-18T14:48:11  <xorAxAx> in a wiki
2006-06-18T14:48:23  <Fabi> basically we are taking about taking the wiki with you on the laptop
2006-06-18T14:48:32  <xorAxAx> yes, thats my primary goal :)
2006-06-18T14:48:36  <Fabi> ehmm talking
2006-06-18T14:49:08  <Fabi> and may be sync some common pages over a set of wikis
2006-06-18T18:04:14  <Fabi> is there a reason why users are not logged in after they crated a new account?
2006-06-18T18:04:38  <Fabi> IIRC they were in the past
2006-06-18T20:12:22  <ThomasWaldmann> xorAxAx: it would be also nice if it would do remote syncing of n wiki servers (like 1 for each country)
2006-06-18T20:12:47  <ThomasWaldmann> so each is backup for the other (and maybe load balancing, too)
2006-06-18T21:35:47  <xorAxAx> ThomasWaldmann: as long as the sync graph of all wikis is fixed, that is ok