Ludovic Dubost wrote: I think the proposal was to use the fully qualified name _for technical aspects_.

In pages (and anything that involves user) we should accept all kind of links, and yes, relative should be the default (therefore export / import would work). This is almost the same issue with links to pages in the same space, whose relativity can become problematic on documents copy. The only problem I see now, in this light, is if you wanted to dump the database wiki1 and create a new one (wiki2) from the dump, in which case any backlinks stored as wiki1:space.page will fail. It is indeed a redundancy issue, I'm having second thoughts about the idea... _______________________________________________ devs mailing list devs(a)xwiki.org http://lists.xwiki.org/mailman/listinfo/devs

On Wed, May 6, 2009 at 17:09, Ludovic Dubost <ludovic(a)xwiki.org> wrote: This example as nothing to do with what i suggest since this is user input and not technical generated datas. I'm speaking about internal generated datas, datas you don't find in the export, the link table is not exported etc...

Hi, On Wed, May 6, 2009 at 6:31 PM, Ludovic Dubost <ludovic(a)xwiki.org> wrote: There's maybe something I don't understand here, but isn't it better to have more data and run a filter to remove unnecessary data in specific cases if needed (such as the MySQL backup scenario decribed above) than to have too little data and to get stuck when trying to identify which wiki a given page belongs to? Would writing such a filter be complex? That is, it's easier to have too much data and to filter out the noise than the other way round, isn't it?

Guillaume

Thomas Mortagne a écrit : I tend to think that what I say is true for any data. As Anca points out a mysql backup of the backlinks data with absolute links would not work if you recover it under another db name.

>> We should always store information as "relative". If a document is >> in >> the same wiki then you should not specify the wiki. >> If it is in the same space you should not specify the space. >> >> Imaging you want to export/import a wiki that was in a farm in a >> database names 'wiki1' and you import the pages in a farm where the >> database is names 'wiki2'. Then your import process will have to run >> conversions, or you wiki will be fully broken. >> >> I'm for generalizing "relative" storage if there are places where we >> hard code the wiki name or the space name. >> >> Ludovic >> >> Thomas Mortagne a écrit : >> >>> Hi devs, >>> >>> The real full qualified name of a document is wiki:Space.Page. I >>> think >>> we sould always use only this form for technical purposes like >>> storing >>> in database or old API which does not support DocumentName. >>> >>> pros: >>> - don't need to take care of the context, the information is >>> always right. >>> - if everything is stored in only one form it's way easier to >>> compare >>> things and do requests >>> >>> cons: >>> - old code which does not support this document name form will >>> fail. >>> This could hardly be a good arguments since this form is a >>> officially >>> valid document name but i can't see anything else. >>> >>> Among others, the last use case which make me send this proposal is >>> http://jira.xwiki.org/jira/browse/XWIKI-3754. IMO getLinkedPages >>> return and the link table should only contains fully qualified >>> links, >>> especially since theses lists are supposed to contains uniques >>> documents and we need to know the 3 parts of the document name to >>> make >>> sure of that . Note that a document could contain a link to a >>> document >>> in another wiki or even a link to the same document but written >>> in its >>> full form so a code which support only local form is wrong anyway. >>> >>> WDYT ? >>> >>> Here is my +1. The pros are basically the same that make us >>> introduce >>> DocumentName to not manipulate String anymore in code. >>> >>> -- >>> Thomas Mortagne