User talk:Cryptic C62/Graphs

Test order

As I understand it, there is a lot of parse caching used to accelerate the visual editor. As a side effect, first time someone attempts to visual edit a specific revision may generally take longer than subsequent tries to vedit the same revision. I'm wondering if that biased your results since you apparently reloaded the same pages multiple times. Dragons flight (talk) 05:01, 3 August 2013 (UTC)[reply]

@Dragons flight: When testing, Firefox was always first. Occasionally the Firefox load time would be significantly higher than the others, presumably due to the effect you described above. When this happened, I just ran the test again and recorded the lower value. --Cryptic C62 · Talk 15:13, 3 August 2013 (UTC)[reply]

Opera

I see you have done tests with Opera 15. Now that VE is working with Opera 12* it might be worth repeating the tests with that. Opera 15 is a Chrome clone so would tend to follow your results for Chrome. Opera 12 and earlier, the original Opera based on the Presto engine, is likely to be used by the majority of Opera users for a very long time as most will not be upgrading to Opera 15 for a very long time. Opera 12.16 can still be downloaded from ftp://ftp.opera.com/pub/opera/win/1216/ When I briefly tried VE here the other day in Opera (12.14) I got similar load times to what you are showing. Until load times get down to 1-2 seconds I cannot see VE getting accepted. Dsergeant (talk) 05:35, 3 August 2013 (UTC)[reply]

@Dsergeant: Excellent suggestion! I will jump on this. --Cryptic C62 · Talk 15:20, 3 August 2013 (UTC)[reply]

Does VE actually function in Opera 12.14 for you? I started testing Opera 12.16, and while VisualEditor does load (slowly), most of the functionality is broken. I'm not sure how useful the load time data would be. "How long does it take to set up a new refrigerator in a house with no electricity?" --Cryptic C62 · Talk 15:42, 3 August 2013 (UTC)[reply]

I only had a very brief play and did not use it for editing. It would not surprise if that a lot of VE doesn't work. Incidentally the release of Opera 15+ is almost a big a disaster as VE is... a totally different browser and not Opera in my view. Dsergeant (talk) 16:49, 3 August 2013 (UTC)[reply]

Model

I think you'll find an appreciably better fit with a quadratic polynomial rather than an exponential. Also, could you post a table of your times? Dragons flight (talk) 17:52, 3 August 2013 (UTC)[reply]

Times and correlation coefficients posted. The degree 2 polynomial model was an improvement only for Opera. --Cryptic C62 · Talk 19:58, 4 August 2013 (UTC)[reply]

Thanks for the times, but I think you labeled the columns incorrectly. The "Chrome" column shows missing values, but those seem to appear on the "Safari" graph. One or more of these labels is presumably wrong. Dragons flight (talk) 00:56, 5 August 2013 (UTC)[reply]

Good catch, thanks mate. I mistakenly pulled the browser order from the graph rather than the data. --Cryptic C62 · Talk 01:01, 5 August 2013 (UTC)[reply]

Re-running these tests?

Hey,

Would you be interested in re-running these tests? We've believe that we've made some significant improvements in VisualEditor's performance over the past few months, and it would be interesting to see if they're reflected in your real-world (ish) experience.

Jdforrester (WMF) (talk) 01:08, 4 February 2014 (UTC)[reply]