Wikipedia talk:VisualEditor/2014 RFC

Initial feedback from the VE team

Hey, and thank you for the notification on Wikipedia talk:VisualEditor asking for feedback.

Recently we have been asked a few times by community members when the English Wikipedia will have VisualEditor switched back on by default for all users (except, naturally, for those who've opted out). This came up the last two times at the monthly office hours I hold on IRC, as well as in several community discussions. This complex question is something that is worth having a community-led discussion about, and I'm happy to have this proposal begin such a discussion.

That said, I think this draft, though a good start, would be even more helpful it it were adjusted before being put out for community input.

Firstly, I think that this is a little premature in approach. We've made a number of significant changes to VE, but the most significant user experience improvement that we've made in months — the new, simplified template dialog — hasn't gone live here yet, and we're still busy adding the new citation toolbar button that we've discussed before.

Similarly, the "background" section should probably give some examples of the things that have changed about VisualEditor in the past seven months. I'd be happy to help here, though obviously you may have your own views as to which changes are the ones to highlight (and ideally the RfC should talk about changes which are actually available on the English Wikipedia for RfC readers to consider). Of course, all features in VisualEditor (and MediaWiki more generally) are works in progress, like Wikipedia, and so drawing a magic "line in the sand" where something is "done" isn't going to really help.

Secondly, I would recommend a slightly different framing around the questions asked. Rather than asking "are we ready to go, yes/no?", I'd suggest focussing on the areas that people feel are necessary for us to move forward, and give us a worklist to help prioritise the issues that matter most to this community. This is touched on in part of question 3, but I think it's sufficiently important that it should be asked straight up; questions about "is it ready?" can then flow from that point.

For example, I know that a lot of community members expressed concern about edits corrupting pages when editors used VisualEditor — from my point of view these have almost entirely been fixed, but if you should still have some examples of bad, wikitext-breaking edits using VisualEditor that should be fixed before it is rolled out more widely, we'd like to know about them. The performance of VisualEditor has been significantly improved since July last year, and a number of pages that were previously un-editable can now be successfully changed using VisualEditor without resorting to wikitext, such as redirects or adding alt text to images.

Finally, there is no question in the RfC as stated about making VisualEditor a default option for existing user accounts (as opposed to new accounts and for logged-out users) – personally, I think that it would be a mistake to expect new users to use VisualEditor (and so come to e.g. the Village Pump with questions) only to find that the experienced users are all unable to answer their questions about it because they do not have it switched on. I'm not sure that having different experiences for new and existing users is something we should encourage.

I hope that this feedback is useful, and I'd be happy to give more details or respond if that'd help.

Jdforrester (WMF) (talk) 20:38, 11 March 2014 (UTC)

Thanks for your feedback. I deliberately included a notice at the top indicating that the discussion is not yet open. In fact, I was actually planning on waiting a few more months before opening it, but if you think sooner is better I'll certainly oblige.

You're right that it would be a good idea to mention some specifics about what's improved; is there a list anywhere of the major changes in the past months?

Your second point is important; I rearranged some of the questions in response. It's probably a good idea to mention some possible areas of improvement to help guide the discussion.

I've added a section accordingly. However, I'm not sure force-changing all user preferences is a good idea; most experienced editors would rather configure their settings by themselves. On the other hand, you're right that having different experiences for new and existing users can be problematic, but plenty of existing users have already selected to use VisualEditor, so is it bad if a few don't use it? (The preferences page says "27,082 users have enabled this feature.")

One thing I'd like to avoid is having an RFC now, getting it shot down, another one in 6 months, that one also failing, repeat indefinitely. That's why I'm spending time and asking for input before opening it; more ideas would be welcome. -- Ypnypn (talk) 21:28, 11 March 2014 (UTC)

I'm going to reiterate an important point from JDF, above. I think that making this a discussion about what the community feels is the level of development necessary for certain policies is going to be more productive than asking whether we should have access X, Y, or Z at this time. Like for me, until the developers have template editing locked down and rigorously tested for several months, I would oppose letting any IPs have access to the visual editor, and would oppose moving to opt-out for account holders as well. To answer the RfC as it is currently written, I would have to do an awful lot of research in order to figure out whether that means I would support or oppose some of the questions. But having it instead from the perspective of being at what point you would support a given level of rollout I think would give a lot better feedback. VanIsaac_WS^cont 14:39, 12 March 2014 (UTC)

@Vanisaac: Thank you. BTW, I'd note that anonymous users have the ability to add ?veaction=edit to pages they're viewing to use VisualEditor right now, mostly for testing purposes; it's currently used about one or two times a day. I'm not sure if this is what's meant in the RfC as a "hidden" ability. Jdforrester (WMF) (talk) 19:31, 12 March 2014 (UTC)

@Ypnypn: Thanks. The main source of changes would probably be Wikipedia:VisualEditor/Updates, though weekly updates about VisualEditor might be good to fill in some details about changes. I'd point out that most experienced editors have never changed a preference, and that doesn't mean they're not experienced. Assuming that all 70 million accounts will eventually get around to actively choosing whether to opt-in or not to VisualEditor seems optimistic. :-) Jdforrester (WMF) (talk) 19:31, 12 March 2014 (UTC)

Thanks for the links; I'll try to add some info tonight. I'd actually like some stats about how many experienced editors change their preferences, not including the 69 million who are long gone. (I think there are only ~100,000 active accounts.) -- Ypnypn (talk) 20:04, 12 March 2014 (UTC)

Tables

I don't think it's even worth having this discussion until VE supports tables and some method of viewing HTML comments. Those two have been on the "to-do" list forever and are part of the minimum viable product. It's hard to find an article larger than a stub that doesn't include a table. The comment problem just leads to fights: an editor has every right to expect that if he puts in a comment that says "Don't change this value without discussion on the talk page because ..." that any other editor that changes the value read the comment.—Kww(talk) 16:10, 7 June 2014 (UTC)

Thanks for sharing your opinion. I was interested in your assertion that "It's hard to find an article larger than a stub that doesn't include a table." I looked at the first ten items in Category:Start-Class medicine articles, and found zero tables (and a couple of infoboxes, which render as tables, but aren't edited that way). Then I thought it might be a quirk of the one subject, so I checked Category:Start-Class Academic Journal articles, and found zero tables. Category:Start-Class Anglicanism articles gave me two succession boxes, but no tables. Category:Start-Class Australia articles had two (20%) with tables. Category:Start-Class aviation articles had one. Overall, in my search, only 6% of start-class articles contain tables.

I agree with you about the HTML comment problem. Last I heard, it was waiting on some design work. Whatamidoing (WMF) (talk) 21:30, 7 June 2014 (UTC)

OK, "larger than start class". It doesn't render my point invalid: an editor that can't edit tables is an editor that cannot edit a substantial subset of Wikipedia articles, including many of the most active, such as those devoted to music and television.—Kww(talk) 23:47, 7 June 2014 (UTC)

It would be interesting (to me) to know what proportion of articles actually contain tables. I believe that many of the pop culture articles use template-creating tables rather than straightforward tables. At the moment, you can change the content, but the ability to change the structure is limited. I've seen editors say that they can change the number of rows, but I believe that's all. A proper table editing tool is going to be some months away, last I heard. Whatamidoing (WMF) (talk) 01:23, 8 June 2014 (UTC)

• It doesn't support tables? Holy crap, how did this thing ever get into testing in the first place? That is absolutely unconscionable that this thing would get approved with such defects in basic functionality. VanIsaac_WS^cont 02:18, 8 June 2014 (UTC)

Doesn't seem ready for prime time yet

I haven't used VE for some time, and haven't been following the development progress, but thought I would give it a quick look-see and drop by with my take on the current status.

• Looking at the Recent changes log, I found just nine of the most recent 5000 edits (0.18%) edit-filter tagged as VE edits (Tag: VisualEditor)
• I just got as far as the third VE edit before I found a problem: diff Maybe that's not fair, as it's the kind of mistake that might be made with a conventional edit.
• But, at this point I think the focus should be on selling more editors on the idea of opting in to use the editor (again). I'd say that a minimum of 5 – 10 percent of edits should be using VE before it's made the default. If you can't even get voluntary usage to half a percent of edits (25 of 5,000) then it's probably not ready to even be sold yet. Wbm1058 (talk) 18:13, 7 June 2014 (UTC)

Wbm1058, I think that percentage of users, rather than percentage of edits, would be a more sensible count. What do you think of that measurement? Whatamidoing (WMF) (talk) 21:44, 7 June 2014 (UTC)

Hmmm, that's like the difference between the US House of Reps. and US Senate... in the "House of Editors" (one-edit-one-vote), the most prolific editors have more influence, while in the "Editor Senate", each editor has one vote, whether they've made 1 edit or 10,000 edits, so "small editors" have more representation there. It's an interesting thought; I'd split the difference, how do we get the percentage of editors who have made at least one edit using VE in the past month? Wbm1058 (talk) 22:26, 7 June 2014 (UTC)

Have you seen the stats at http://ee-dashboard.wmflabs.org/dashboards/enwiki-metrics before? That gives "Daily unique editors" (NB that it double-counts people who use both VisualEditor and the wikitext editor on the same day). For a month-long numbers, there may be another method, but mine is going to be asking for a database query when someone in Analytics has a little time free. It may take a while to get an answer.

Given the experience at other Wikipedias, I think that expecting highly active editors to make 10% of their edits with VisualEditor is unrealistic, even if it were turned on as default. The reason for this is partly because of the way we count edits: every edit using "undo" is counted as an edit made with the "wikitext editor" (because it is, or at least could be, if you made some change other than a straight undo). Also, about 20% of editors are unable to VisualEditor even if they wanted to, due to a lack of Javascript, using an elderly browser, using Internet Explorer, etc. Quite a few highly experienced editors don't want to use VisualEditor, or any kind of rich text editor. These are people who find wikitext positively attractive. Others simply can't be bothered to learn the new system (and why should they have to?).

It would make more sense to compare it to what's happening at communities where VisualEditor is available by default already, and use that as your baseline for what's possible when it is default-available to everyone. For example, if you look at a Wikipedia like French, you'll see that 8.8% of all non-bot edits use VisualEditor today. User stats show that it sometimes runs a bit higher there, but you get more active editors on the weekend. At the Swedish Wikipedia, today's use is only 6.6% of all non-bot edits. The last time I checked (and if memory serves), only about 2% of currently active editors at the English Wikipedia have opted in. This is actually pretty good for anything that's opt-in, but with 1/50th of the editors having the option on their screens, you should realistically expect 1/50th of the edits that we see in places where almost all editors have access to it. That works out to something approximately like the 0.18% that you found today, which, by the way, would be consistent with the ~10% use that we saw here last August when almost everyone had access to it.

Having inflicted all of that on you, I agree with your overall point: I'd much rather see more people opting in for testing than turning it on for everyone all at once right now. Whatamidoing (WMF) (talk) 22:59, 7 June 2014 (UTC)

No, I hadn't seen that stats page before, thanks for the link. You do have a wealth of data on which to base decisions. Wbm1058 (talk) 15:42, 9 June 2014 (UTC)

Testing report

I'm one of the minority of editors who has opted in, perhaps in a show of "moral support", so sees the edit (beta) tab, but virtually never has used it for the past several months. I realize I may not be part of the target market for VE, but nonetheless, I'll try using it for all of my edits, and report on each, until I tire of that. Wbm1058 (talk) 15:42, 9 June 2014 (UTC)

• Sorry, I gave up on this attempt. Could not figure out how to remove this template and comment with VE, after some good-faith fiddling with it. Wbm1058 (talk) 15:42, 9 June 2014 (UTC)
• The built-in disambiguator is nifty, and I figured out that clicking on the trash can removed the link. This is serviceable, but I still believe that the religious refusal to support [[ ]] in VE is seriously holding back its acceptance. Wbm1058 (talk) 15:56, 9 June 2014 (UTC)
• After some fiddling with it, I figured out that the 105th Belmont Stakes redirect can be created using the options dropdown. But the refusal to let me do it any other way, leaves me feeling the attitude, "This is France, you should not speak English here" (even though we understand the language). What's the problem with supporting multiple methods of accomplishing a task? Wbm1058 (talk) 16:22, 9 June 2014 (UTC)
• So why isn't VE smart enough to know that this edit should never be allowed? Very annoying that it refuses valid edit attempts, yet allows invalid attempts. I thought I was on the way to deleting the template by clicking on the trash can, but the "save page" button was greyed out and I couldn't figure out what the magic trick was that would un-grey it so I could click on it. Sorry, I will need to fix this one with a conventional edit. Wbm1058 (talk) 16:29, 9 June 2014 (UTC)
• Created Camp Cedars Equestrian. Easier the second time around. Wbm1058 (talk) 16:46, 9 June 2014 (UTC)
• I figured out how to change categories with minimal effort (although WP:HOTCAT is more efficient for this). Why didn't my edit summary automatically show the category change, as Hotcat does? Wbm1058 (talk) 16:46, 9 June 2014 (UTC)
• Somewhat amusing that in this diff, VE lets me put a [[wikilink]] in my edit summary, even though it refuses to do that in the edit itself! Again, the disambiguator is the best new feature I've seen since last I looked at VE. Wbm1058 (talk) 17:03, 9 June 2014 (UTC)
• The "clear formatting" dropdown was helpful for making this edit. Wbm1058 (talk) 17:20, 9 June 2014 (UTC)
• I see attempts to create a redirect like this all the time—this needs constant patrolling. An easy, more intuitive way to accomplish the task begs for a VE solution, i.e. a way to create the BBL 2013-14 redirect while editing the 2013–14 Big Bash League season article. Wbm1058 (talk) 17:32, 9 June 2014 (UTC)

OK, 26 of the last 5,000 edits were made using VE, and that only includes the last 2 that I made. Done testing for now. Wbm1058 (talk) 17:42, 9 June 2014 (UTC)

This is a great list; thanks for posting it. I'm going to move it over to the main feedback page (which has a lot more watchers), so that more people will get to see it. Whatamidoing (WMF) (talk) 19:15, 10 June 2014 (UTC)

Curious about how VE might impact article creation for new editors

First, let me say that the new reference tool is a signficant improvement. I still use Reftools (and an early version of it at that) because of the autofill logic, but I would probably teach a new editor to add references with VisualEditor than with the code editor at this point, if that were an option.

I have always imagined that one of key goals of VisualEditor was helping new editors overcome the terrible learning curve that Wikisyntax can be. Having been putting in a lot of hours at Articles for Creation recently, there's little question to me that Wikisyntax does create a pretty signficant barrier to entry. Hardly surprising, of course, but it's worth repeating, and it started getting me wondering -- just how would the benefits and drawbacks of dumping new editors into VE rather than the code editor play out for new editors who begin, as many do, with the desire of creating a new article?

Creating a new article as a new editor is generally an impossible job, of course, we do a (redacted) job of explaining to new editors what our standards for inclusion are. Over and above issues of wording and copyright, most of the "work" at AfC involves notability, and in nearly every case, notability revolves around references and sourcing. Here, wikicode creates an additional barrier where the terrain is already rocky.

Where I'm going with this: I wish it were possible soon (it isn't, now, on ENWIKI) to do a couple hundred editor test of new editors doing new AfC article creation with VE. Follow them through a real article creation process. Figure out if it helps or hurts article survival, the amount of work necessary for handholding the new editor through the process, and so on. VE does work in Draft space, so that's good (and it's been helpful, I find a lot of punctuation and wikilink cleanup of accepted submissions to be easier with VE than code editor now, and use it regularly for that task.)

I'm sure there are a few blocking technical issues before we could attempt this, most critically, AfC makes use of hidden comments for part of its work, there might need to be a little assistance somewhere there. I'm not sure that tables are blockers, actually, for a new article creation, someone can create a table-like filmography without tables and have that edited for style after AfC.

Because references are such a central part of the AfC process on ENWIKI, and because difficulties at AfC translate directly into new editor retention issues, I understand that even a trial will generate some concern. But, for precisely the same reasons, I would think it could be quite valuable research. *ponder* --j⚛e decker^talk 17:23, 9 June 2014 (UTC)

Hi j⚛e decker,

If you wanted to know the answer to that question, it should be possible to get the WMF Analytics department to set up a proper A/B test on new accounts (half of new accounts get VisualEditor, half don't). We'd have to wait for the hidden HTML comments problem to get resolved, and it would be good for the regulars at AFC to have some familiarity with VisualEditor (so that any advice they provide is relevant), but it would take a while to get everything planned with Analytics anyway. Whatamidoing (WMF) (talk) 19:18, 10 June 2014 (UTC)