Last modified: 2010-06-21 21:11:47 UTC
Dear friends, Please take a look at http://meta.wikimedia.org/wiki/User:Gangleri/remarks#InterW iki_link_translation and http://en.wikipedia.org/wiki/User:Gangleri/remarks#proper_l ink_translation . Please see the differences. Thanks! Regards Reinhardt
More about anchors: Part I: Just for your information. Would be to much to fix. a) Sections and subsection titels can look very odd. See [[:en:Wikipedia talk:Categorization#Poll_about_Category:Whatever.7C.2A_.2C_.7C.28space.29.2C _.7C.21|en:Wikipedia talk:Categorization#Poll about Category:Whatever|* , | (space), |!]] and look how the link to the anchor was made. b) Tray to create a section / subsection and use an internal link [[foo]] and /or an interwiki link [[:zn:foo]] or an external link [http://www.foo.int/ foo] in the title. Save the section / subsection and look at the url in your browser. c) Tray to use '''font''' ''styles'' and / or "quotes" and / or 'marks' as title. Tray all the special characters !%&?*# and so on. Part B: At [[:is:Notandi:Gangleri/test/Snið:Flokkatré]] there are two "'''anchors'''" <nowiki><font id="Landafraedi" /> and <font id="Island" /></nowiki>. They work only because in both cases special characters as "æð" in "Landafræði" and "Í" in "Ísland" have been replaced. I would be happy, if the issue with the special characters and ( and ) could be solved. Best regards ~~~~
See also [[bug 845]] and [[test:User:Gangleri/tests/anchors]]. Regards Reinhardt
Here's a step-by-step procedure to illustrate at least one of the bugs: 1. Open this URL: http://test.wikipedia.org/w/index.php?title=User:Gangleri/tests/anchors&action=edit§ion=17 . 2. Click "save page" (need not change any text in the page). 3. Expected: URL loaded: http://test.wikipedia.org/wiki/User:Gangleri/tests/anchors#edit_here Actual: URL loaded: http://test.wikipedia.org/wiki/User:Gangleri/tests/anchors#.5B.5Bedit.5D.5D_here Similarly for http://test.wikipedia.org/w/index.php?title=User:Gangleri/tests/anchors&action=edit§ion=18 the browser goes to http://test.wikipedia.org/wiki/User:Gangleri/tests/anchors#here_.27.27.27too.27.27.27 instead of http://test.wikipedia.org/wiki/User:Gangleri/tests/anchors#here_too.
Additional remark(s): Spaces (also multiple subsequent spaces) are translated correctly in an anchor of an interwiki link. See some examples (and an link to real "[[en:Pathology|pathological]]" anchor) at http://test.wikipedia.org/w/index.php?title=User%3AGangleri% 2Ftests%2Fanchors&diff=0&oldid=4348 Regards Reinhardt
If I understand this report correctly, the precise problem can be summarised as follows (for reference, this paragraph would have been an ideal initial comment for this bug report): When typing an internal link to a section, you can use "special" characters like "(", ")", etc and they will be converted to the "escaped" forms like ".28" and ".29" (e.g. "[[#foo (bar)]]" is equivalent to "[[#foo_.28bar.29]]"). However, InterWiki links don't undergo this transformation, resulting in invalid links (e.g. "[[fr:Foo#foo (bar)]]" will create a link to .../wiki/foo#foo_%28bar%29 not .../wiki/foo#foo_.28bar.29) I think this is basically a matter of how much we assume to know about the target of an interwiki link - many interwiki links are to non-MediaWiki sites, which won't follow these special escaping rules. One possibility would be to escape the link differently depending whether the interwiki prefix is marked in the database as "local" or not (i.e. assume that everything 'local' is a MediaWiki, and everything else isn't). At first sight, it would seem logical to make incoming URLs with "wrong" characters in behave as though they had them escaped, but this is awkward because the wiki can't manipulate the address requested other than telling the browser to reload as though it was a different page (it's entirely up to the browser to scroll the page based on text after the "#"). To reply to some other comments: (In reply to comment #1) > a) Sections and subsection titels can look very odd. There's really not a lot that can be done about that; escaping of some kind is inevitable unless we simply restrict what characters headings can contain. However, when working correctly, you should never have to type these escaped forms, so it shouldn't matter. > b) Tray to create a section / subsection and use an internal link [[foo]] > and /or an interwiki link [[:zn:foo]] or an external link > [http://www.foo.int/ foo] in the title. Save the section / subsection and > look at the url in your browser. That's an unrelated bug: the "add section" code is guessing wrong what the anchor will be on the final page. It might be worth filing another bug for. > At [[:is:Notandi:Gangleri/test/Snið:Flokkatré]] there are > two "'''anchors'''" <nowiki><font id="Landafraedi" /> and <font > id="Island" /></nowiki>. They work only because in both cases special > characters as "æð" in "Landafræði" and "Í" in "Ísland" have been replaced. I'm not sure what you mean here; in what sense did they "not work" before this change? Can you come up with a simplified test-case to demonstrate? If you just mean that interwiki links weren't generated correctly, as with the main issue in this report, then that is to be expected, for exactly the same reasons.
At least _part_ of the issue here is the difference between the way that the parser creates anchors from headings (for the purpose of linking from the TOC), and the way that in-page targets are handled in links (e.g. [[PAGE NAME#heading]]. The code in parser.php (function formatHeading) seems to do more work to ensure that the anchor can be linked to. In linker.php, function makeKnownLinkObj, the code is a bit simpler (less robust?), and a very similar bit of code exists in editPage.php, function sectionAnchor. Given the relative opacity of the code in parser.php, I'm not sure how practical it would be to pull the code there into a function visible from linker and editPage. Something along this line is probably needed.
(In reply to comment #5) > > b) Tray to create a section / subsection and use an internal link [[foo]] > > and /or an interwiki link [[:zn:foo]] or an external link > > [http://www.foo.int/ foo] in the title. Save the section / subsection and > > look at the url in your browser. > > That's an unrelated bug: the "add section" code is guessing wrong what the > anchor will be on the final page. It might be worth filing another bug for. Filed as bug 1860.
bug 2381: "enhancement request: traying to achive copy and past anchors" suggests a workaround to handle some / many of the described anchors
At some point in the last 5 years and 10 days, the parser started anchor-escaping interwiki links in the same way as local links. There are other bugs with anchor link generation, see bug 18431 and bug 5019 et.al., but I think the issue identified in comment 5 has been solved, so I'm closing this bug.
