On Sat, Oct 30, 2010 at 11:39 PM, Magnus Manske
<magnusmanske(a)googlemail.com> wrote:
It should be comparatively simple
to write a more generic "transfer" parent class, which then would have
derived classes for each of these sites; a simple method (e.g. for an
image page on flickr, give me the URL of the most high-res file) might
be sufficient for specialization.
Good idea, I will take that into account
Second, the special case of transfering from other
MediaWiki sites.
This includes all Wiki(m|p)edia projects, as I do in [2] with >400K
uses (and [3], when it works), but also WikiTravel, and basically any
other MediaWiki installation where a license can be determined. While
it might seem to be easy to implement this, as we are more familiar
with the site behaviour,
there is no API for image metadata in
MediaWiki,
There's a bug for that.
<https://bugzilla.wikimedia.org/show_bug.cgi?id=25624>
and transcoding the wikitext correctly, from all
projects
and languages, can be a real b***h, as countless more-or-less botched
transfers from my bot show. Alternatives would be parsing the HTML
(lossy), or putting more weight on the user to check for correctness.
Even if you do not chose to implement any of these
transfer options
initially, I believe you should code with these as further additions
in mind. IMHO it would be a real shame to "waste" such an opportunity
on flickr alone.
I agree, but whenever I'm going to do this, I am going to make a
generic interface for sites that can output metadata in a structured
way. This means that Flickr, Picassa and others can be easily
supported, but that MediaWiki will only be supported when the
aforementioned metadata is supported by MediaWiki. I believe that
eventually MediaWiki will support such features, but we will by then
be well into the 1.20s.
Reagards,
Bryan