Daniel,
I suggest there isn't a need for "data objects (that) will be cached in the client wiki (i.e. wikipedia)". In the transclusion approach, nothing at all is cached by the client wiki except the infobox's HTML, within the squid cache as has been outlined. IOW, I'm still wondering *why* data objects must be retrieved and then cached by anyone, in the first place. And, so that I fully understand the wikidata approach, isn't it actually true that the API calls that you say are NOT occurring "during the render process (that) would be very scary", are indeed executed during every purge of a wikipage?
If, as you imply, transclusion meets wikidata's functional requirements, then would it still be necessary to require every wikipedia to install the client/server API (aka the "wikidata client")? What is so compelling about the client/server approach? You say "Keeping the control over the formatting in the client wiki seems desirable to me" overlooks the fact that normal wiki rules-of-the-road apply within the wikidata environment also, where authors certainly should be able to exert "control", likely even more so, over infobox content & styling.
Thanks - john
On 10.07.2012 14:08, Daniel Kinzler wrote:
On 10.07.2012 16:47,
jmcclure@hypergrove.comwrote:
I am concerned about the performance
impact of every wikipedia calling an API for each property that it wishes to format as content in pages' infoboxes, as I understand is the design the project is pursuing.
No, that's not the case. The data
objects will be cached in the client wiki
(i.e. wikipedia) and be
loaded into memory once for any page that uses them.
Doing API calls
during the render process would be very scary.
Could you please
explain to this community why it's technically superior to field a client/server API rather than transclusion, e.g., {{wikidata:en:infobox:Thomas Jefferson}} It seems more stable a design to format the infobox on wikidata, and then simply transclude the result.
Keeping the control over the formatting in the client wiki
seems desirable to
me, though I also see the appeal of a central
repository for infobox templates.
We could just have both though, just
like for images: use the local version if
it exists, otherwise use
thetemplate form a central repo (which may be on the
wikidata site or
somewhere else).
-- daniel