On Sat, Jun 15, 2013 at 6:09 AM, MZMcBride z@mzmcbride.com wrote:

Greg Grossmeier wrote:

...

• On Tuesday VisualEditor team will enable an A/B test, where half of

new accounts created on English Wikipedia will get VisualEditor enabled by default. This is to test performance and features before the larger rollout in July.

As I commented at https://bugzilla.wikimedia.org/49604, the (apparent) lack of an easy means of opting out of this experiment and the increased frequency of severe VisualEditor-related bugs being reported recently both make it appear that an A/B test of this nature would be premature and potentially very damaging. Dirty diffs, inadvertent section removals, etc. are still common when using VisualEditor. Are we really expecting our newest users to be able to spot and correct these issues?

I'm also quite surprised to see that VisualEditor could be activated by default for some new accounts.

I tried again VisualEditor on frwiki, and I don't see how it could be effectively used by new users : * VE is still very limited : for example, not being able to edit template is clearly a big limitation. How a new user will react when he tries to edit an article for a first time and see many parts that he can't edit ? * VE is still doing dodgy things in some situations : how a new user can deal with modifications done by VE without his knowledge ?

Currently, I think that VE should only be activated voluntarily, so that people know what they are doing, are prepared to fix incorrect modifications made by VE, and can easily work outside of VE to overcome its current limitations. I think that new users would quickly get a bad first impression if using VE.

For me, VE is currently a good start for a more user friendly editor, but as it still lacks a few important features and still has a few bugs, it should stay in opt-in mode and clearly not in opt-out mode.

Nico

I disagree with the last three messages: the visual editor is bound to be painful when enabled; if the team has established that they're in a stage where they are done with the knowns in some areas but need more testing on the field (and feedback) to discover more, they must be allowed to, otherwise they'll surely lose time, not working on important issues they couldn't predict. I'm not able to make such an assessment, but I read today a discussion on it.wiki where it was noted how only 132 edits were made with the visual editor. It's impossible, for the community (or anyone) on any given wiki, to gain any confidence in the tool as long as its usage is at such negligible levels. VE is useless for power users, so we can only try with newbies. I don't know if new accounts are the best choice, maybe it should be triggered only after the 10th edit (we know at this point the survival rate is much better); but this can be changed any time, I guess.

Nemo

Can you explain how to set a user preference based on the number of edits a new user has? Has the VisualEditor team set up a script to roll back (i.e., undo) this change to the user preferences of new users if this experiment doesn't work well? Personally, I think it would be a little crazy to press forward with this experiment on June 19 as planned.

Based on what I have seen the few times I have fooled around in Visual Editor (not sure if I actually saved any edits or just said screw this and went with the regular editor), I also think that this is a bit premature.

If we need additional feedback, rather than automatically enabling it for half of the new editors, or even half of the new editors with X number of edits, we could provide a notice to new editors with X number of edits that asks them if they would like to try Visual Editor. We could provide a notice at the top of Visual Editor to allow them to easily change the preference back as well.

While this doesn't solve the template issue (and the clean-up that will inevitable come of that), it might work as a solution to the getting new users to test VE issue.

Thank you, Derric Atzrott

Hi,

The interest of the VE team is not the only one to take into account I think... The impact on the new wikipedia editors should be a more important parameter in my opinion.

I tried VE a few times, and clearly think it's not yet in a situation where it could be rolled out to unexperienced users : * VE is still very limited in what you can do with it (no templates, no references, ...). What will be the reaction of a new user when he sees that he can't edit some parts of the article ? * VE is still quite buggy (adding nowiki tags, deleting references, modifying templates, ...). While it's not a problem with users that opted-in for testing, it's quite different for users that don't even know what VE is. * Beta testers made a few suggestions for enhancements that would be quite helpful for editors (like being able to choose between VE and wikitext when editing a given section and not globally, ...)

Why do you want to rush a forced test on new users when VE is not yet a stable, fully functional product ?

You mentionned the low number of edits with VE currently. I think it's low because of the problems mentionned above, not because of a lack of testers. I saw several users do like me: try it, see that many editions can't be made or end up with side effects, report the problems, and use again the wikitext waiting for the problems to be solved in a next version. I do believe than once VE is stable and has more features, people will start to use it more widely.

Has there been any analysis done to foresee the impact on new users that would have VE enabled by default ? Like taking a few hours or a day of modifications on enwiki, keeping only the modifications made by users registered in the last few days, and try to do the exact same modification with VE : * What percentage of modifications could be achieved with the current set of features available in VE ? * What percentage of modifications would have been done without undesired side effects ? That would give an idea on how many new users would run into problems with VE (for me, they are very low, but I'm not a new user). With the current version of VE, I believe both those percentages will be low, implying many new users will have problems.

Nico

On Mon, Jun 17, 2013 at 12:12 PM, Federico Leva (Nemo) nemowiki@gmail.comwrote:

The fact that there are known issues doesn't mean that finding new, unknown issues will slow down the work on the known; it's up to the team to decide what sort and what amount of feedback they'll be able/need to process (and to adjust if they were wrong).

Gradually enabling a feature is not an "experiment" on some poor victims, it's a normal development strategy (as opposed to sudden revolutions/waterfalls on the wikis). I still don't see any indication of why it should raise the end net harm of the VE development on the wikis.

I don't know how to enable the preference at some point of users' lifecycle; probably, in the same way you do it for half new users. A hook I assume, it was mentioned in some Echo and enotif bugs.

Nemo

______________________________**_________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/**mailman/listinfo/wikitech-lhttps://lists.wikimedia.org/mailman/listinfo/wikitech-l

On Mon, Jun 17, 2013 at 5:42 PM, Nicolas Vervelle nvervelle@gmail.comwrote:

I tried VE a few times, and clearly think it's not yet in a situation where it could be rolled out to unexperienced users :

• VE is still very limited in what you can do with it (no templates, no

references, ...). What will be the reaction of a new user when he sees that he can't edit some parts of the article ?

I would always double check the latest version before we talk about VE's limitations -- the feature set is evolving pretty rapidly right now, to the credit of the team. For instance, VE now supports adding and editing references and templates. (This weekend on English Wikipedia, I was able to pretty smoothly add navigation box templates and page protection templates to articles.)