Noted, thanks. Perhaps I'll get around to discussing performance benchmarks down the road, and as you say, perhaps I'm overthinking this.
Pine
On Mon, Jun 15, 2015 at 8:49 AM, Tilman Bayer tbayer@wikimedia.org wrote:
On Fri, Jun 12, 2015 at 10:06 PM, Pine W wiki.pine@gmail.com wrote:
I agree that there is value in continuity, but remember that Wikipedia articles change over time, so unless someone is using a specific rev for measuring every time that they make a change to how the page renders,
then
there is likely to be at least some unreliability in the measurement. Technical factors like bandwidth and geolocation may also be involved in skewing the validity of comparisons.
An astute observation that shows your deep knowledge about wiki technology and web performance measurement. However, I'm pretty certain that for this particular project, Facebook's HHVM engineers tested the same revision of the article, and knew how isolate/separate varying bandwith.
For most citations, there appears to be a manually updated list here:
https://en.wikipedia.org/wiki/Wikipedia:Articles_with_the_most_references
I wasn't able to find a list of articles with the most templates,
although
there are a few articles where the template expansion depth limit is exceeded.
As mentioned in the blog post that we want to socialize here (http://hhvm.com/blog/9293/lockdown-results-and-hhvm-performance ), "MediaWiki was benchmarked using the Barack Obama page from Wikipedia, as was recommended by an engineer from Wikimedia foundation as representative of their load." What you describe sounds more like a recipe to find outliers, not examples that are reasonably representative. Also, as Jeremy mentioned, the Obama article has a bit of a tradition among MediaWiki developers as an example of a somewhat complicated but popular article, also in other context than performance - see eg.
https://commons.wikimedia.org/wiki/File:Wikipedia_Android_app_screenshot_aft... or search for "Obama" on Phabricator. I think it's a straightforward choice as the head of state of the country where Wikipedia is hosted, and not really politically charged at that. I'm usually all for crafting social media messages carefully and avoiding gaffes, but in this case we may be overthinking things a little.
Perhaps we should take the discussion of how best to measure page
rendering
performance to Wikitech. Would that be ok with you?
I don't want to stop you from educating the performance engineers at Facebook and WMF about the wrongness of their ways, but we should indeed get this social media message out soon, and I assume it would take Facebook a while to re-run their study anyway. In general, you may be interested in https://phabricator.wikimedia.org/T97378#1285776 .
Pine
On Fri, Jun 12, 2015 at 9:43 AM, Jeremy Baron jeremy@tuxmachine.com
wrote:
On Jun 12, 2015 12:41, "Pine W" wiki.pine@gmail.com wrote:
In terms of byte size, that article isn't even in the top 250. See
https://en.m.wikipedia.org/w/index.php?title=Special:LongPages&redirect=...
And by template count? Or cite template count?
Also there's some value to continuity. We've been using the Obama
example
for years.
-Jeremy
Social-media mailing list Social-media@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/social-media
Social-media mailing list Social-media@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/social-media
-- Tilman Bayer Senior Analyst Wikimedia Foundation IRC (Freenode): HaeB
Social-media mailing list Social-media@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/social-media