On Fri, Sep 4, 2009 at 4:56 PM, jeroen De Dauw<jeroen_dedauw(a)yahoo.com> wrote:
After discussing with Ævar, we agreed that using the
Maps extension 
instead of SlippyMap to display OSM maps with OL is probably a good idea,
assuming this is doable of course, which I think it is.
I am able and willing to put quite some time into
making the needed changes
to Maps. Before I can really start though, it would be helpful if I had the
most up to date code of SlippyMap, since a lot of it's current code can be
re-used. Is this the one committed to the SVN trunk? (I've downloaded that
code, but it's missing at least one file and giving other errors as well.)
I run the code from SVN (although with this patch:
shouldn't matter) and it works just fine. What file is missing / what
errors do you get?
It would also be nice if the people who are working on
this end of the
mapping effort give me a poke, so efforts can be coordinated, and no one
goes off to create duplicate or redundant functionality.
If anyone has objections to using Maps instead of SlippyMap, please voice
them, so we can discuss the advantages and disadvantages of using Maps.
To provide a bit of context Jeroen contacted me on IRC recently about
this and I talked to Yaron (the other main author of Extension:Maps)
about this as well.
I don't care what piece of software we use for embedding maps into MW
pages as long as it works. So if you're willing to modify the Maps
extension so that it works for our purposes that's just great. One
less thing for me to worry about.
While I don't care /what/ software we use whatever we end up using has
to be suitable for production deployment on Wikimedia sites for our
purposes. Which means, among other things:
* Making it fast
I'm vaguely worried about the size of the Extension:Maps codebase
compared to what we actually have to use. But then again I'm not that
familiar with it and you said on IRC that it was all pluggable so you
could just enable the parts you need.
* Making it secure
I fixed up a bunch of potential XSS injections in the SlippyMap
codebase by being really paranoid about what input it accepts (see
I also added parser tests (which Extension:Maps lacks) to make sure
this all works.
Perhaps you can reuse some of the parsertests work for Maps?
* Making sure it doesn't bite us later
The SlippyMap extension has a really limited featureset currently.
Anything we enable we're going to have to support in the future (& has
to be scalable). So Maps will have to be configured to expose a really
limited featureset (e.g. no semantic stuff, or API queries to Yahoo).
These are as I see it the most important functionality
* Migrate code from SlippyMap to Maps, mainly the static map handling
Rather than copying the huge pile of PHP code that calculates a
bbox/scale from lat/lon/zoom it would probably be more worthwhile to
work on this: https://bugzilla.wikimedia.org/show_bug.cgi?id=19861
But in any case I made the lat/lon/zoom -> bbox/scale code isolated
from the rest so feel free to copy/paste until we have something
better on the backend.
* Add the ability to display maps without any markers
on them. Currently
maps has several parser functions, like display_point, which will display an
error when you don't provide any coordinates instead of displaying an empty
map. (Should be easy to do though.)
Regarding markers I was going to remove the marker= support in the
SlippyMap extension, because:
* It's limited. We want some sort of syntax to support multiple markers
* In order to support anything like markers/kml layers/whatever we're
going to have to support that in OpenLayers *and* on the static
* Optimize the code. Currenly Maps supports OpenLayers
as mapping service,
and allows you to display maps from a variety of mapping services. On top of
that, the OL JS is far from optimal. Therefore I think it's best to simply
add a new service to Maps (this is done via a hook-like system build in
Maps), and create optimized handling for only OL+ OSM there, based on
current SlippyMap JS.
* Optionally (probably not required in the first
version) add image-as-layer
capabilities, allowing users to zoom and pan around high resolution images.
That's probably better done as a seperate project for File: pages as
OpenLayers gets wider use within MediaWiki.
Anything mission critical I'm missing there?
Not that I can see.