Map Of London

KM-F reports that at some locations in the UK, she has come across machines running IE8 or below (at least, that's what I'm assuming based on the described symptoms). I'm not sure whether that's the only browser available on those machines -- surely not -- but IE8 is the newest version of IE available for XP.

IE8 cannot handle application/xhtml+xml; it offers to download and save the file rather than display it. Our redesign site is pure XHTML5 and is served with the correct mime type. I've therefore created a trap in the controller.xql like this:

if (matches(request:get-header('user-agent'), 'MSIE\s+[1-8]\.')) then
    <dispatch xmlns="http://exist.sourceforge.net/NS/exist">
    	<forward url="obsolete_browser.xql">
    	</forward>
    </dispatch>

This has to come before everything else, of course. The obsolete_browser.xql page is just a simple plain HTML page that suggests updating IE if possible, or installing another browser (links provided).

Luckily I still have an XP VM which I was able to use to test this, although it took lots of updating because it hadn't been run for ages. In the process I also created a little library for showing the HTTP headers eXist sees, for help in debugging.

One day this problem will go away. Surely.

TL pointed out that when an editorial note appears in a field such as an author field, and the author field is used in a TOC table where the rest of the document itself is not available, the note number link was attempting to pop up the note but showing a not-found message because of course the note is not available in the TOC page.

I've worked around this by catching situations like this in the XSLT; when such a note appears in content which is inside an XHTML table cell (this is the structure produced by the mdtList.xql TOC-generator), the note is processed differently, and does not turn into a link. The note content is still accessible with a mouseover because its content is in the title attribute anyway. An extra class changes the cursor to "help" instead of "pointer".

Fixed many instances of the mol: protocol being used where there is no actual @xml:id to point to; updated the documentation to clarify that in these cases, a simple link to the .htm page should be used. Working through the bad links systematically. Some will have to wait until after the redesign is launched before they can be fixed (since there are some pages on the new site which are not generated in the old codebase).

I've rewritten the consistency checks so that they work in a modular way by AJAX. You choose an individual check and retrieve only results for that. We can run checks against the file we're working on by putting its xml:id into the text box, or we can run checks against the whole db (which takes a long time). If you run the first check against the whole db (be patient), there are nearly 80 bad links that we need to fix -- I fixed about ten already. Some of these are links to site pages which actually work, because they're converted to whatever.htm and the page happens to be there, although there's no @xml:id (these should actually point directly to the html page). However, the ones that are most urgent are links to the short capital @xml:ids which don't exist such as this:

There is one link  using @corresp, @ref or @target pointing to "WRIG3", but there is no item with the @xml:id "WRIG3".

We should start working through those and making sure they're all fixed by the time we get close to the launch.

Note to self: in the process of doing this, I figured out how to have a single XQuery library that uses different serialization options depending on the nature of the query. The base page is constructed if there's no checkType parameter, and serialized as HTML5, but if there's a checkType param then it's an AJAX call and I serialize it as XML, by putting this into the local function which generates the output:

  let $method := util:declare-option('exist:serialize', 'method=xml'),
  $media-type := util:declare-option('exist:serialize', 'media-type=application/xml'),
...

I've just finished writing an XQuery unit that does some basic checking for consistency and issues with xml:ids and linking in the database. Anyone can check a file that they're working on (assuming they have upload privileges and you can put it into the database) like this:

consistency_checks.htm?id=MIDD17

where the bit at the end is the xml:id of the document. For those of us working on a more general level, you can run the file without a parameter and it will check the whole db. It takes a little while to do that.

There are currently 72 issues it knows about (I already fixed some straightforward ones while I was writing the library.) I'm going to be adding some more tests to this as we go forward. In particular, I want to flag occasions where the <ref> element is used instead of <name> to link to a person.

The A-Z index page has now been modified so that it works in two ways: the default is that it shows only documents, or items from documents, whose status is 'final', 'published', 'stub' or 'generated' (in other words, documents that we want people to see on the site). However, there's also a requirement that encoders and contributors be able to see the complete list of ids, so if you add the parameter "show=all" you'll see everything. The old URL ids.htm has now been aliased so that it points at the "show=all" variant.

Populated all the info documents with more catRefs. Also fixed a rendering bug when following a search result link, and began testing of CloudTest Lite, which JJ has suggested we use. Might be handy for performance tuning in the run-up to the launch.

<term> elements with @corresp pointing to an entry in the glossary now turn into links which pop up the gloss (if there is one).

I've substantially updated the search interface so that it now allows checkbox-selection of multiple document types, rather than a one-only dropdown list, and also so that it allows the configuration of different search pages with different pre-configured sets of document types to choose from. The main search is configured for four document types, and there's a customized search page at serach_documentation.htm which specializes in documentation subtypes. The search will also handle the submission of a name parameter, and when you do that, it handles things differently, searching first for any matching id for that name, then searching on that id. There's no interface for making use of this feature, though, as yet; it's a holdover from the original search system on the PHP webapp, where the only thing you could actually search for was a name. But I don't know that we need this sort of thing now, because any of the major name types provide facilities for finding all the documents which mention them/.

Finished the old map interface, with working categories from the new taxonomy. Met with JJ and TL, and came away with a new list of stuff to fix, especially around the Search. Then implemented a feature to retrieve the list of documents a place is mentioned in, to show at the head of a location article, as well as sprucing up name highlighting, which had gotten lost in previous XSLT work. Now if you go to one of these linked articles, the name of the place whose article you came from is highlighted.