> Hm? I thought I wrote the rules out already... What is missing, to > be exact?
You're right, I've found them at the end of the page. >> Regarding backwards compatibility, I'd recommend ignoring old special >> cases (where existing) in favor of allowing for simple rules for 3.0+. > I thought that's what I said in the proposal :) > The only thing really to wonder about is the english plurals - > setting. There's some annoying magic in ReferenceManager to make > sure that "TestPage" and "TestPages" are considered same objects. >> Especially the hidden platform dependencies, which are obviously >> existing in the file-based provider up to 2.8 should be eliminated. > These are more or less errors, not really features. Ok, that's fine then :-) However, I think the proposal is lacking some clear words on under which conditions a link is case sensitive or not. To me it seems as if this is somehow dependent of and implied in the tryBeautified switch... >> * The simple way: Implement some fallback rules in wiki page lookup in >> order not to break existing links. If feasible, the corresponding >> code >> should be marked as being for this purpose only. >> Problem: What if some old link isn't valid any more and therefore >> gets permanently redirected by this code? Then, someone creates a >> new page whose name is just this old link's name. This would break >> the old link without touching it directly. > Could you elaborate on that? The current recommendation to create > links (2.8) is to simply put them inside brackets. For example, > [This is a page] creates a link to a wikipage called "This is a > page". In 2.4 and previously, this would actually create a link to > "ThisIsAPage". > Our current implementation checks first for the existence of "This is > a page", and then for backwards compatibility reasons, > "ThisIsAPage". However, internally, these pages are not considered > the same, so it is possible to create "This is a page" even when > "ThisIsAPage" exists, and the new page will then take precedence. > Because *most* of the links at jspwiki.org are these backwards > compatible forms, I don't think we can throw it away. However, what > we can do is to make it a configuration item. Sorry... Having read your mail and thinking about it again, I realized that I've been mixing this up with the current case insensitivity "errors" when running 2.8 on a windows machine. Another thing I didn't mention in my last mail is the wiki:title. Which benefits do we have from maintaining this extra property, which seems to be always automatically changed along with the page name? Regards Florian
smime.p7s
Description: S/MIME Cryptographic Signature
