Wikipedia talk:STiki/Archive 23

This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

Archive 20

Archive 21

Archive 22

→

Fixing STiki's hanging/session issues

Greetings folks. As a classification session of my own tonight hopefully demonstrates, I am very much interested in solving the issues everyone is having with STiki hanging or dropping sessions. I can see that tool usage is down across the board -- and this speaks to many people having difficulty. This will lead to tool abandonment, decreased adoption, and generally bad news for our brand/philosophy moving forward.

I used STiki from 12:29AM (UTC) onward tonight, for at least 50 minutes, making probably close to 100 edits (combining reverts, AGFs, warnings, etc.). Only once about 7 minutes in did the program hang (on an AGF revert). On my Mac I was able to (Command+Q) to quickly kill STiki's window, and restarted and had no issue moving forward. It appears all of my edits made it onto my WP contributions. That one hang didn't give me the strong trend I was looking for. I deliberately restarted several times (not prompted by an error) in order to see if that would screw things up.

I need folks to be as detailed as possible in making crash reports. I assume this never happens after "pass" or "innocent" actions? and is therefore descriptive of some interaction with the WMF servers, or POST actions in general. Is anyone else on a Mac? I think I am running JDK 1.7 in my development environment. I'm on the East Coast of the United States, FWIW, running on a decent speed connection. I am an admin and have rollback rights. I have a globally unified login but haven't dealt with the third-party authentication stuff the WMF plans to roll out. Unfortunately, firewall restrictions at work make it impossible for me to check in during normal (US east coast) business hours if someone is having difficulty at those times. Try to run STiki run from the terminal -- if you know what that means. A hanging application is unlikely to output anything to the terminal, but in other situations it might. Let's solve this together. Thanks, West.andrew.g (talk) 01:15, 7 July 2016 (UTC)

Andrew, similar to yourself I have limitations in testing (firewall, OS, etc. - I only use Stiki with WinXP). I can tell that crashing does not depend on the action (reverting or passing diffs) and user status (I am an admin and many others above are not). I check Stiki about twice a day these weeks, last time being 3 hours ago, and the crashing behavior persists. Possibly I connect through another Wikimedia server. Maybe you can simulate that by using an open proxy. Materialscientist (talk) 01:27, 7 July 2016 (UTC)

I'm currently on an ancient Windows (XP), and a relatively slow connection. I've had two errors today; a session timeout when I went for an AGF revert (after a span of several minutes when STiki was running in the background, but I wasn't using it); and a hang when I went for an "innocent" classification, which I had to kill with task manager. If necessary, I will post the error message the next time it appears. Vanamonde93 (talk) 05:30, 7 July 2016 (UTC)

Knowing things can hang after "innocent" classifications is useful. I would be delighted to find this is something wrong with STiki's server communications that is well within our power to fix. I can also write automated scripts that keep issuing innocent classifications (say, from the end of the queue no one will ever reach) until things break, speeding up debugging. Thanks, West.andrew.g (talk) 12:58, 7 July 2016 (UTC)

Andrew (a naive comment, not to forget). Stiki often crashes because it overloads CPU. On my PCs the (long-term) overload can be ca. 98% on WinXP, and is unbearable, but it is limited to 50% on Windows 8, and I can keep running Stiki with it. Normally the CPU load by Stiki should be much lower. I hope you didn't have that in your test runs (meaning that you could go on simply because your PC easily handles CPU overloads). Materialscientist (talk) 22:29, 7 July 2016 (UTC)

Materialscientist, you may have something there; each time that STiki has hung on me in the past few days, my CPU usage has gone to 100%. Still, I was using STiki before it hung and usage was well below that; so I'm not sure in which direction the causality runs; does it hang because it's using all the CPU capability, or does it use 100% CPU because it's hanging? I don't have the know-how to figure it out. Oh, and the last time it hung was on a "pass." Vanamonde93 (talk) 06:27, 8 July 2016 (UTC)

This is useful. For STiki CPU usage to hang near 100% one or more of the threads is probably stuck in an infinite loop. This speaks less to the networking issues we have been speculating about. Last time there was a CPU-usage issue, I think one of the parser routines was to blame, perhaps because the WMF diff formatting changed ever so slightly. I'll watch CPU usage on my machine (which is admittedly quite powerful) next time I fire up a session. Thanks, West.andrew.g (talk) 13:48, 8 July 2016 (UTC)

For me, the problem isn't that STiki crashes or hangs - crashes never happen, and the very occasional hanging resolves itself after a few seconds. This is despite CPU usage never breaching 20% on Win7. My problem is that STiki often in a session never actually reverts any edits, despite giving off the appearance of doing so - yet the 'last revert' box remains empty with "No prior revert: No warning data". Repeatedly restarting will get it to work eventually, but it's generally an exercise in frustration. Just got my first working session in several weeks, so am pretty happy about that. I don't think the problem lies at my end (i.e. STiki is actually doing something), because after a restart, revisions classified as 'Vandalism, 'GFR' or 'Innocent' no longer show up in the queue, but revisions classified as 'Pass' still do. Seems STiki is classifying correctly but not reverting - this is coming from a computer newbie though, so take everything I say with a large grain of salt! Satellizer el Bridget ^(Talk) 08:56, 8 July 2016 (UTC)

Stiki is Ok with me since Monday - I presume because Wikimedia sysops changed/refreshed something in the software/database. This happened many times before, so I expect the same Stiki blackouts will recur in the future. Materialscientist (talk) 08:32, 12 July 2016 (UTC)

I had a couple of session terminations today, so it doesn't seem to have been completely fixed. Or it's something on my end. Vanamonde93 (talk) 10:00, 12 July 2016 (UTC)

Ah yes. There were two different problems before: 1) Stiki doesn't revert, 2) Stiki overloads CPU and crashes. It seems 2) is gone and 1) remains. Materialscientist (talk) 10:24, 12 July 2016 (UTC)

STiki not able to run on my Windows 10 PC

I upgraded my Windows 7 PC to Windows 10 yesterday. I downloaded STiki today and unzipped it. When I open STiki, the file asks "How do you want to open this file?". I want to ask how to run STiki on Windows 10. I have also downloaded and installed Java and JDK. Even then also, STiki isn't running. Mr. Smart ℒION ⋠☎️✍⋡ 12:34, 12 July 2016 (UTC)

It sounds like your computer doesn't know that it wants to open STiki with Java. I'm not familiar with Windows 10. When it asks "How do you want to open this file?" is it possible to point it in the direction of Java? Yaris678 (talk) 17:09, 13 July 2016 (UTC)

Yes, it's possible to point it in the direction of Java. However, I could also point it to the wrong Java file. What's the exact Java file to point in STiki? Mr. Smart ℒION ⋠☎️✍⋡ 17:46, 13 July 2016 (UTC)

All you need is the JRE in this case, FYI. I doubt that alone will solve the problem at hand, though. I think for most Windows user's the "How do you want to open this file?" dialogue never appears, because a JAR is quite universally understood to be a Java executable file extension. This speaks to possible problems in the JRE/JDK installation. West.andrew.g (talk) 22:37, 19 July 2016 (UTC)

Changing the warning style to #Month# #Year# format?

Sorry if this question has been asked before, but I can't seem to find any documentation concerning the markup that STiki uses (let me know if I'm overlooking it somewhere). How can I have STiki do Twinkle style #Month# #Year# instead of "Regarding your recent edits to #a#"? And if the heading exists already, will STiki automatically put the warning under it? Thanks! – FenixFeather ^(talk)^(Contribs) 05:18, 23 July 2016 (UTC)

Never mind, I found it in the help pane! I guess it's not possible. Follow up question: is there a way to change the default values of STiki edit comments? – FenixFeather ^(talk)^(Contribs) 23:47, 23 July 2016 (UTC)

@FenixFeather: In your comments are you talking about: (1) the section title of the warning? (2) the edit comment associated with the warning? or (3) the edit comment associated with the revert? #1 is not editable for good reason. The date structure STiki uses is standardized across multiple tools (CBNG, Huggle, STiki) so they know when/where/how to auto-increment warning levels. I don't think we expose #2 to users, but it might be hidden in STiki's configuration file (I can't recall). #3 is edit-able in the bottom-left "comments" panel. What are you trying to do? Thanks, West.andrew.g (talk) 15:52, 26 July 2016 (UTC)

@West.andrew.g: Thanks for the explanations! Yeah, I did figure out that the vandalism warnings follow the style of everyone else, so I'm not worried about that anymore. With my second comment, I was talking about #3. I have gotten used to just editing in the box and then hitting default when moving on to the next edit. I figured it would be useful if I could change the default to automatically have parentheses or something after the default message so I can just fill in the reason, but that's just a minor inconvenience. – FenixFeather ^(talk)^(Contribs) 17:30, 26 July 2016 (UTC)

Request to use STiki

I applied a while back, and I was told to apply to WP:CVUA. I have graduated (verify) and now have more edits. I'd like to get started with live after the academy. Thanks, ProgrammingGeek (Page! • Talk! • Contribs!) 22:35, 23 July 2016 (UTC)

Congratulations! A 100 edits in article-space, though, is not quite enough as far as my standards go. My advice: carry on the good work as before, pick up a few 100 more edits in main space, and we give you a warm welcome here. Anyone else? Super48paul (talk) 15:13, 25 July 2016 (UTC)

I would make the same recommendation. You've made a solid start, keep it up. In addition to more than a 100 article-space edits, though, I would recommend spending at least a little while trying to build content. You don't have to do anything major, but right now you don't have more than four edits to any single article. Some experience with trying to create content will help you better determine when others are being constructive, and when they are not. Regards, Vanamonde (talk) 04:38, 26 July 2016 (UTC)

I don't mind the lack of content work so much. It's not really all that relevant to STiki. Your work here so far has been fantastic, so another, solid, 100 countervandalism edits and I would be more than happy to second a request of yours. Good luck, and feel free to ask me if you've got questions. Omni Flames (talk) 09:45, 26 July 2016 (UTC)

I disagree that content creation is needed to fight vandalism. I had no content creation experience when I started fighting vandalism on Wikipedia, and I wasn't a disaster, was I? Wikipedia:Requests for permissions/Rollback states that applicants "must have a measurable track record of counter-vandalism", not about content creation. The WP:CVUA exists to train vandal fighters, and creating content, IMHO, does little to help. —k6ka 🍁 (Talk · Contributions) 12:52, 26 July 2016 (UTC)

Hmm, I didn't make myself clear, it would seem. I am not asking for GAs or DYKs or anything of that sort. All I am saying is this; working on an article for a little bit will help with recognizing some of the unconstructive stuff, not just vandalism, that we see; and I'm sharing this from experience, because in my view, I got better at using STiki as I got more experienced with content. Perhaps that's not true for everybody. Regards, Vanamonde (talk) 13:27, 26 July 2016 (UTC)

The edit-count tool is down right now, which is usually my approach for vetting these things. When I look at someone's experience, I mainly look for anti-vandal or other patrolling experience, knowledge of the warning system, and civility in talk-page dealings. Indeed, a breadth of experience would make someone a more refined STiki user. However, given the volume of vandalism, all we really need is someone with the good sense to press "pass" when ambiguity arises. I also feel that we told this user to complete CVUA in order to get access -- they did what was asked -- and now we are raising the bar again. I'm inclined to approve on that basis alone, though I'd also like to see more edits. West.andrew.g (talk) 16:10, 26 July 2016 (UTC)

@West.andrew.g: obviously use your judgement, but completing CVUA proves I can fight vandals, and with a few STiki edits I'll be able to apply for rollbacker. I could get this permission manually editing, but it would take longer to find the vandalism in the first place. ProgrammingGeek (Page! • Talk! • Contribs!) 08:27, 27 July 2016 (UTC)

West.andrew.g, since the edit count tool isn't working for you, they've made around 140 Twinkle edits, and a total of around 100 to mainspace. Most of the Twinkle edits appear to be countervandalism, although they've also welcomed some new users. They seem to be warning new users correctly from what I can find. Omni Flames (talk) 08:55, 27 July 2016 (UTC)

@ProgrammingGeek: Speaking as an editor who got STiki access by reaching the 1000 edit mark, don't worry about the difficulty of manually fighting vandalism. It's actually not as hard as you think, and you learn quite a bit along the way. The recent changes page helps a lot. Also, beware of Wikipedia:Hat_collecting. Why do you need rollback so badly and so quickly? – FenixFeather ^(talk)^(Contribs) 08:57, 27 July 2016 (UTC)

@FenixFeather: The reason for rollback is to also use Huggle. ProgrammingGeek (Page! • Talk! • Contribs!) 10:38, 27 July 2016 (UTC)

Done -- @ProgrammingGeek: The user has been able to have a sensible dialogue about this with us. I have assumed good faith and granted access. I will spot-check his/her contributions in the coming days, and I hope he/she will exercise caution given this discussion. In the worst case, STiki really just enables dumb people to do dumb stuff quicker. In the future I will try to be more clear that CVUA completion alone might not meet the threshold. Good luck and happy reverting. West.andrew.g (talk) 19:51, 27 July 2016 (UTC)

@West.andrew.g: Hello, I'm ProgrammingGeek's trainer. Thanks for accepting the request, as I believe PG will do well. Also, I might've suggested for PG to wait a bit, but it seems like it will work out at the end. Finally, even after PG graduated I am still monitoring some of his edits. I'll do the same and try to check his STiki edits even more. Dat Guy^Talk_Contribs 10:28, 28 July 2016 (UTC)

Can't run on my Max OS X Yosemite

I try signing in but I keep getting an error message as follows: "Error in the user login interface, likely caused by a network error. Exception thrown to terminate thread." Any thoughts? I am currently in Mexico until August, but I don't think that would do it... Any help would be much appreciated. Thank you! -Pax Verbum 02:17, 26 July 2016 (UTC)

@Pax85: STiki worked fine for me under Yosemite, and now El Capitan. That message is the same one you would get if you incorrectly enter your username or password, so double check that (I know because I just noticed and re-wrote the error message in source code last night). STiki will have issues if your network firewalls port 3306 (mysql), but we do that connectivity check before you even have the opportunity to login, and you'd get a pop-up dialogue if that were the case. So.... double-check your credentials and let me know if the problem continues even after a couple restarts? If there was a systematic login problem, I'd expect to see more complaints here. Thanks, West.andrew.g (talk) 16:19, 26 July 2016 (UTC)

@West.andrew.g: That's what I thought at first too, but after double-checking for three days, still no luck. As far as I know there are no odd network settings at my location, and I've tried on two different networks here in Mexico. -Pax Verbum 17:07, 26 July 2016 (UTC)

@Pax85: Does a diff load in the main STiki panel if you wait long enough without logging in? Just as a sanity check, you are able to log-off and log back into Wikipedia from a normal browser using those credentials? Do you know your version of Java? Are you able to launch the JAR from the terminal, and if so, what gets printed to the terminal window when you see the failure dialogue in the GUI? Thanks, West.andrew.g (talk) 19:40, 27 July 2016 (UTC)

@West.andrew.g: Well, I know I would fail the sanity check, but I am indeed able to log in from a normal browser. As far as the others, I will let you know later tonight. Sorry for the delay; I've been a bit busy. -Pax Verbum 18:07, 30 July 2016 (UTC)

@West.andrew.g: So I was checking everything out, and on a whim decided to make sure Java was updated. After the update, all is well now. Thank you! -Pax Verbum 00:15, 2 August 2016 (UTC)

CHANGELOG for 2014-12-01 release (+ subsequent bug fixes and ERs)

Greetings STiki-ers... We have a substantial release this iteration, fulfilling many months of feature requests and minor bug discoveries. This is the first version in a while whose release was not forced by an emergency circumstance (i.e., unannounced WMF change), and I am excited to get out meaningful features that demonstrate my commitment to the tool and the anti-damage community. This version brings us:

The "Options" menu now has a setting that allows clients to be notified via pop-up if an AIV post has been lodged (T#018; T#046).
Null diffs should no longer be displayed (T#038). This was a bug introduced when STiki enabled custom color options for users.
Users will not be displayed their own edits from the queues, as this is a conflict of interest (T#043).
More AGF messages were added, including some pulled in from a GitHub merge (our first community code contribution!). Some category labels were pre-pended to help better manage this growing list. The code copy was also synced with community copy at: WP:STiki/Good-faith-revert_messages.
The metadata panel content now has "mouseover" support. In particular, this should permit the full display of comments which are truncated because they are too lengthy to fit in the fixed-size placeholder.
A "small numerical edits" filter has been created. Sometimes monotonous and difficult to casually assess, this filter can allow users to skip edits where "only a single numerical value is changed, and the magnitude of that change is (0.5 < change < 2.0)" (T#049).
Additional watchlisting options (reverts/warns/both/preferences/never) are now accessible via a drop-down menu in the login panel (T#042).
Additional user/article-talk links added to metadata panel (T#036).
Removed reference to "editing anonymously" in login dialog (T#044).
The "don't template the regulars" option/dialog was re-focused such that it is now labeled: "warn if reverting regular" (T#050).
A classification button was added to support reversion and immediate "4im" issuance (assuming warning criteria are met and escalation wouldn't already result in an AIV report)(T#041)
There have been reports of user's edit session's dropping, resulting in editing while logged out. This results in privacy loss and community confusion. This is believed to be a bug on the WMF server-side. However, using the "assert=true" check, STiki can now at least detect the error state -- it pops a dialog and forces a restart (T#048).
Due to a WMF API change, we were forced to change the way user edit counts are performed. The WMF changed the way pagination is done, breaking our earlier parser in some cases. This code was used to determine STiki permissions (>1000 NS0 edits), "don't template the regulars" logic, and non-critical aspects of milestone reporting.
Minor improvements to warning increment logic; in particular, better handling when duplicate section headers exist.
Documentation changes in support of above
- 2014-12-01: An additional space character in some of the pre-formatted AGF messages was causing them to be parsed in a non-conventional boxed style with monospace font, characteristic of the "pre" environment. This was fixed and and a new file was uploaded under the same filename.
- 2014-12-12: Something seems to have changed with the way Mediawiki parses new line instructions, causing section headers to not be displayed, and causing warnings to run together. STiki has changed line breaks on talk pages messages from "/r" (carriage return) to "/n" (newline).
- 2015-06-13 EMERGENCY RELEASE: The WMF switch to support only HTTPS caused widespread breakage due to hard-coded HTTP API calls. Paths were updated and the version was marked as a required upgrade. This also pushed T#051 to production which addressed a bug with the persistence of custom AGF messages.
- 2015-07-02 EMERGENCY RELEASE: Mediawiki is changing the way pagination/continuation is handled for API queries that produce many results. This decision breaks existing code, but curiously, legacy functionality is still supported if an extra parameter is passed (if this sounds dumb, it is). The old methodology will be switched off sometime on 2015-JUL-02. Internal to STiki this primarily affects/affected the permission check of "1000+ edits in NS0". This new version fixes that issue; users who who rely on that entry criteria will likely require an immediate upgrade.
- 2016-01-22 EMERGENCY RELEASE: Updates to the MW SessionManager conflicted with how STiki was constructing its session cookies. An invalid session meant that rollback/edit tokens were not validating, and no reverts/edits could be made. This fundamentally broke earlier versions, thus a forced updated was issued.
- 2016-02-12 EMERGENCY RELEASE: Authentication changes are upcoming (we'll try to integrate them before they break us, for once!) and some method deprecation warnings were added to the API login process accordingly. These messages broke our expected response format, parsing, and all of login as a result. Forced update issued.

I have tested the code, but as we develop features around infrequent use-cases (4im, AIV, etc.) it becomes increasingly difficult to verify correctness in the wild. Your continued monitoring, innovation, and support is greatly appreciated. West.andrew.g (talk) 03:46, 1 December 2014 (UTC)

Suggested feature - Thank button

On rare occasion, STiki pops up an edit that was a revision of vandalism or some other clearly beneficial contribution. I'd like to suggest that a "Thank" button be added so that the STiki user can WP:THANK the user who made the edit. Just an idea that would promote WikiLove and would appear to be a useful addition imho. Thanks! (Please {{re}} me in reply) EvergreenFir (talk) 06:57, 3 August 2016 (UTC)

hello Evergreen: we already can thank an editor by pressing the thanks button in the diff screen. Incorporating such a button in STkiki would only burden the tool - which in general should be avoided. Anyone else? Super48paul (talk) 08:22, 3 August 2016 (UTC)

I'll have to disagree with Super48paul here. Thanking someone currently requires you to leave STiki and use a browser to access the actual diff, where you can then thank the editor. Having to do this is a pain and it would be far better to just have a built-in feature. Also, by that logic, we shouldn't have a "revert" option either, because you can press the revert button in the diff screen. Omni Flames (talk) 08:52, 3 August 2016 (UTC)

I would support EvergreenFir's request. It would indeed be helpful to be able to thank a user from the STiki interface. While this is a tool usually used to revert vandalism, if we can do so while being as friendly and encouraging to newbies as possible, this would be a good thing. Vanamonde (talk) 09:27, 3 August 2016 (UTC)

@Vanamonde: But you cannot identify a newbie from the STiki interface, you will have to click through to the diff screen anyway (assuming that you installed appropriate popups).. Super48paul (talk) 08:43, 4 August 2016 (UTC)

@Super48paul — I disagree. If it's a constructive edit, I'd "thank" regardless. ProgrammingGeek (Page! • Talk! • Contribs!) 15:59, 4 August 2016 (UTC)

@EvergreenFir: I'd been thinking the same thing. Would love this feature. ProgrammingGeek (Page! • Talk! • Contribs!) 21:52, 3 August 2016 (UTC)

I will add to feature table with "Medium-Low" priority. Given my limited cycles with STiki, I am always afraid of mission creep. In an ideal world, we wouldn't want anyone to ever use the "thank" button, because ideally STiki would show only suspicious edits with a high probability of vandalism. Should this be a button in the main classification panel? Does "thank" imply "innocent" in the sense they should be tied together (in the way "4im" implies "vandalism") ? West.andrew.g (talk) 15:28, 5 August 2016 (UTC)

I do believe that even if you have a software that is only flagging suspicious edits, a thank button would be appropriate. Here's why: a substantial number of the edits that come up on my screen are an IP making a small spelling or wording change. Frequently, those are of the test/vandalism variety: sometimes, they are an IP going out of their way to fix a typo or grammatical error. Assuming that these IPs are generally not regular editors, a "thank" button that simultaneously marks the edit as innocent would encourage participation here, at little to no cost to the STiki user, IMO. Vanamonde (talk) 17:00, 5 August 2016 (UTC)

The API functionality is easy to implement. Let's talk about how/where it should appear in the interface. West.andrew.g (talk) 19:42, 6 August 2016 (UTC)

Maybe it should be like the "4im" button by vandalism, and potentially a check-box under the AGF revert message box. ProgrammingGeek (Page! • Talk! • Contribs!) 15:33, 7 August 2016 (UTC)

PG's suggestion makes sense to me: I'd want it to be a one-click feature, but introducing it with the minimum of clutter would be ideal: so a "4im" equivalent on the "innocent" button would be great. Vanamonde (talk) 17:33, 7 August 2016 (UTC)

I agree with the "4im equivalent" idea. I can't see how the check box idea would work. Yaris678 (talk) 18:32, 7 August 2016 (UTC)

When I get back to my computer I'll make a visual on where it would be located. I didn't explain well. ProgrammingGeek (Page! • Talk! • Contribs!) 11:34, 8 August 2016 (UTC)

Improved GUI layout attempt

Hey everyone. When switching to the Mac/Apple platform a little over a year ago, I was surprised to find the GUI layout wasn't as clean as the Windows/Linux machine it was designed on (Java is supposed to be cross-platform, right?). It was work-able, but not quite how I designed it. I've finally got around to making some tweaks. Now, I don't know if I've screwed up the Windows design in the process; or how universal the fix is across different OS/versions/resolutions.

If anyone wants to help, can you screenshot how this new version looks on your machine: TWEAKED GUI VERSION? Thanks, West.andrew.g (talk) 18:14, 6 August 2016 (UTC)

I do not know if the remark is relevant - but though the interface works, I still (sometimes) encounter the (well-known) problem of pressing the good faith revert button and being thrown out. Super48paul (talk) 08:33, 7 August 2016 (UTC)

Yeah, I continue to troubleshoot those issues that are proving very difficult to troubleshoot. STiki is seeing a lot of new users, and the number of daily classifications has markedly increased, so seemingly its not something affecting the entire user-base, at least not at painful rates. I continue to put debugging code in place, but I'm able to go for hours some nights without issue. I've decided to start working on other improvements to the tool, and in the course of testing those features, see if I can encounter some of the bad behavior you've described. West.andrew.g (talk) 14:46, 8 August 2016 (UTC)

For me, using Windows 7, it looks fine when full screen, but if I reduce the height of the window slightly, it takes the height from the "Watchlist options" drop down menu, which makes it look odd. Click on the screen shot on the right to zoom in and see what I mean. Yaris678 (talk) 08:58, 7 August 2016 (UTC)

Acknowledged. Looks pretty good otherwise. The "watchlist options" drop-down does a similar thing on my Mac machine. It seems really dumb to start shrinking a component when there is still dynamic vertical spacing "glue" in the panel [shrugs]; maybe I need to enforce some kind of minimal size. West.andrew.g (talk) 15:04, 8 August 2016 (UTC)

Not sure about enforcing a minimum size. Would that mean that if I have a low-res screen I can't you STiki? Sadly, I don't know anything about how Java GUIs decide which bits to shrink first. Yaris678 (talk) 15:13, 8 August 2016 (UTC)

I meant enforcing a minimum size only on the height of the "drop down" menu. Java GUI construction is mind-numbing in that every single component has a "minimum", "preferred", and "maximum" size. For some simpler components, these don't need to be set, and the defaults generally suffice (seemingly not with drop down menus). The space between components is a dynamic "glue" which is the first thing to shrink as the window shrinks. As components get nested inside components, these matters, along with their respective layout managers (you can organize components according to axes, grids, compass positions, etc.), become increasingly complex in their interaction. West.andrew.g (talk) 15:54, 8 August 2016 (UTC)

I see. Yes. Setting a minimum size for the dropdown menu sounds like a good fix. Yaris678 (talk) 10:47, 9 August 2016 (UTC)

STiki not recognising an edit conflict

I'm pretty sure this edit shouldn't have happened. I edited the article manually and then pressed "good faith revert"... and it still did the revert.

I was using the slightly modified version of STiki linked above (STiki_2016_08_06).

Yaris678 (talk) 07:16, 9 August 2016 (UTC)

No related changes in the slightly modified version. Let me know if this happens again and I'll deeply investigate. Otherwise, the conflict logic has been rock-solid for a long time. Moreover, my statistics feed just got turned off (I knew it was deprecated), so I need to spend my free cycles getting the WP:5000 and related reports back in working order. Thanks, West.andrew.g (talk) 14:54, 9 August 2016 (UTC)

Level-1 warning issued despite level-2 being present

Apologies if this has been addressed already, but STiki delivered a level-1 warning when a level-2 warning was already present (see [1]). I'm not sure why it didn't increment to a level-3 warning in this case. The edit being reverted was [2]. Perhaps it's because the level-2 warning was about OR? Thought I should point out the issue. Cheers. EvergreenFir (talk) 06:42, 10 August 2016 (UTC)

Behavior was "as expected" here. There is not shortage of templated warning messages that one can escalate. If someone gets chastised for one thing, I don't think we should escalate the penalty for doing something orthogonal (which they may be unaware of). I believe this follows the standard of other anti-vandalism tools (Huggle, CBNG). Relatedly, this is why one shouldn't get too eccentric and use rare warning templates. West.andrew.g (talk) 17:30, 11 August 2016 (UTC)

Okay. Wanted to make sure it was intention. Clue bot, for example, typically escalates afaik. I used a precise template to address the specific issue, assuming good faith that this was the only issue and that the acts were more vandalism. Again, I'm newer to stiki so I'll defer to experienced opinions. Just wanted to make sure it was meant to do that. Cheers! EvergreenFir (talk) 03:01, 12 August 2016 (UTC)

Vandalism outside articles

Stiki should show us possible vandalism on user pages, user talk pages, Wikipedia pages as help pages, policy pages.

In Huggle I reverted vandalism on WP:AN. --Marvellous Spider-Man (talk) 07:17, 12 August 2016 (UTC)

@Marvellous Spider-Man: It already does that, is there anything that led you to believe it doesn't show such vandalism? Omni Flames (talk) 12:22, 12 August 2016 (UTC)

@Omni Flames: STiki doesn't show edits outside of the main namespace. What Marvellous Spider-Man is referring to is having STiki show edits from other namespaces instead of just the main one, like Huggle does. —k6ka 🍁 (Talk · Contributions) 12:39, 12 August 2016 (UTC)

Hi guys. I agree that this would be a useful addition. For some namespaces, the machine learning might need to be repeated on a different data set. For others, I suspect the current assessment criteria - based on machine learning in the main name space - will be sufficient. Help and Portal are two obvious examples of these.

This was discussed a while ago and logged at T#028. See #Known Bugs and Feature Requests above.

I think the main thing stopping this happening is just lack of time for Andrew to make the necessary changes.

Yaris678 (talk) 12:59, 12 August 2016 (UTC)

Yeah. The models could potentially be very different given conversational style and typical edit structure. It wouldn't be hard to enable the current model to classify in new namespaces -- but we run the risk of diluting the pool and lowering overall hit rates. These days STiki users are doing about 1000 reviews a day, and the average hit-rate (vandalism + AGF) has been nearing 50%. Clearly we haven't run out of vandalism in the (arguably more important) main namespace. The few cycles I have these days are dedicated to bug fixing, and when possible, making STiki users more efficient over an already rich pool of vandalism. Does CBNG score pages outside the main namespace? West.andrew.g (talk) 15:12, 12 August 2016 (UTC)

Non-English scripts

Non-English words written in Asian language are not visible. They appear as square boxes on Stiki window. --Marvellous Spider-Man (talk) 15:24, 10 August 2016 (UTC)

I'll look into it. I feel like I've seen Hebrew render correctly before, but I may be mistaken. West.andrew.g (talk) 17:32, 11 August 2016 (UTC)

I can't be sure... but I think STiki showed the edit I reverted here correctly.

I think it would take someone with good knowledge of Java to answer this question... but could it be that the unicode support on Marvellous Spider-Man's machine is missing. I wouldn't know if it would be down to the physical machine or the Java Virtual Machine. Yaris678 (talk) 12:51, 12 August 2016 (UTC)

[3] may be relevant. I don't really have time to digest it right now. Thanks, West.andrew.g (talk) 15:17, 12 August 2016 (UTC)

High DPI scaling support?

The java app is still usable with a high DPI screen (ability to increase font size helps), but all the buttons and text boxes are really small and hard to read. Is there some way the size of the UI elements can be altered for high DPI screens? – FenixFeather ^(talk)^(Contribs) 07:16, 19 August 2016 (UTC)

@FenixFeather: I will investigate. West.andrew.g (talk) 13:54, 19 August 2016 (UTC)

You're awesome, thank you! – FenixFeather ^(talk)^(Contribs) 14:42, 19 August 2016 (UTC)

Old session drop issues again? / Not editing

I've been having the old session drop issues again today: in fact I haven't been able to make a single revert. Is this just my connection, or other folks experiencing this too? Vanamonde (talk) 17:16, 12 August 2016 (UTC)

Is this happening when you press good-faith revert? i.e. the same problem as Super48paul is having in #Improved GUI layout attempt, above.

I haven't had this problem myself. However, I notice that there seems to be a lot more vandalism in the feed than there used to be. It may be that many users are getting this problem and Stiki is being less used as a result.

I don't know what is causing this problem.

Does the problem still occur with the latest version of STiki, released this morning (last night in the US)?

If the problem continues, it may be worth addressing T#004 (HTTP/PHP communications) to see if this has any effect.

Yaris678 (talk) 08:35, 14 August 2016 (UTC)

@West.andrew.g: I just did like 5-6 reverts in STiki, but none of them were "saved" on-wiki. Bug in new version? (t) Josve05a (c) 05:06, 14 August 2016 (UTC)

I've just tried the new version and it worked for me. Is anyone else getting the problem that Josve reports? Yaris678 (talk) 09:18, 14 August 2016 (UTC)

That's a recurring problem and has been discussed many times on this page. It happens occasionally, but restarting the program usually helps. Sometimes it takes several restarts. Widr (talk) 10:50, 14 August 2016 (UTC)

It would be god if there were some indications in the tool to see my recent edits or something...:/ (t) Josve05a (c) 13:39, 14 August 2016 (UTC)

Look at your contributions! — Gareth Griffith-Jones | The Welsh | Buzzard | 18:36, 14 August 2016 (UTC)

Lol, fair enough. (t) Josve05a (c) 19:13, 14 August 2016 (UTC)

First, let me say that everyone should be using the recently released "2016-08-14" version. As described towards the end of the CHANGELOG, I did identify an issue with rollback tokens that might cause some of these "session dropped" or "session failed" sessions. These bugs are inconsistent and I've been unable to reliably recreate them for testings purposes, so I can't say for sure how much of a "fix" this change has created. If it proves to be effective, I'll make the new version a required update and break the older versions.
Second, ticket T#004 isn't going to be any help here. That only references the communication between a STiki GUI client and my database server at UPenn. The issues we are experiencing here are related to the STiki GUI communicating with the WMF, which is already done over HTTPS.
Third, is the "Last Revert panel" not helpful in detecting when this happens? That is, it doesn't update after a "vandalism" classification if the rollback isn't actually completed, right?

Thanks, West.andrew.g (talk) 17:47, 16 August 2016 (UTC)

I've been getting this today. It happens only after I click AGF for the first time. After that, no click works, no revert on any article takes place. --Rsrikanth05 (talk) 06:58, 17 August 2016 (UTC)

Apologies. I didn't notice a new version came out three days ago. Will update and then see. --Rsrikanth05 (talk) 07:00, 17 August 2016 (UTC)

Happening specifically on the "first AGF click" is helpful and similar to the rollback issue I identified above. I will investigate further. West.andrew.g (talk) 18:28, 17 August 2016 (UTC)

The issue is happening even after the update. It's all OK till I click the AGF button. If AGF is the first edit, then I get the error and have to close it. --Rsrikanth05 (talk) 18:55, 17 August 2016 (UTC)

Yep, I suspect this has something to do with editing/rollback tokens. Because AGF doesn't use the rollback functionality, we might be seeing (similar to what we saw in the now-fixed rollback case) some initial issue in obtaining edit tokens. Let me see if I can test this out. Thanks, West.andrew.g (talk) 01:07, 18 August 2016 (UTC)

I can confirm what Rsrikanth is saying: the problem only comes up on my very first AGF revert action in an editing session (even on the new version). Vanamonde (talk) 10:23, 18 August 2016 (UTC)

@Vanamonde93: @Rsrikanth05: Can you guys try this new version and let me know how it goes? There is a little bit of debugging output if you are able to run from terminal. Anyone else who frequently has session issues, please also give it a try and provide some feedback. Thanks, West.andrew.g (talk) 20:54, 18 August 2016 (UTC)

West.andrew.g: I just tried the new version, and I had the same issue once again: the session drop occurred when I tried to perform my first AGF revert. It isn't an issue for a "vandalism" classification, somehow. Vanamonde (talk) 04:51, 19 August 2016 (UTC)

@Vanamonde93: Are you able to run STiki from a terminal/command-prompt? Does it print anything? West.andrew.g (talk) 13:51, 19 August 2016 (UTC)

'Fraid not, my skills with anything that doesn't have a user interface are exactly zero. It isn't a serious enough problem that it's stopping me from using it, though. Vanamonde (talk) 13:59, 19 August 2016 (UTC)

No no. I just need you to "launch" the GUI from the terminal. What is your operating system? West.andrew.g (talk) 14:06, 19 August 2016 (UTC)

Oh I see. Windows 7. Vanamonde (talk) 17:13, 19 August 2016 (UTC)

(broadly to everyone) Is DOS prompt still a thing? How does a Windows user launch from terminal/command-line/console/prompt? West.andrew.g (talk) 20:23, 19 August 2016 (UTC)

It's not real DOS, just an emulator, but Start --> Run --> CMD takes you to a "DOS" window. See PowerShell. Dbfirs 20:40, 19 August 2016 (UTC)

@Dbfirs: @Vanamonde93: Yeah, one just needs to be able to navigate to the directory where the [STIki_exec_ver.java] is stored and issue "java -jar STiki_exec_ver.java", assuming that java is hooked into the shell environment. Can anyone actually confirm this works in PowerShell? West.andrew.g (talk) 20:57, 19 August 2016 (UTC)

I've just tried that under Windows 7 cmd and STiki ran successfully, but I think Windows 7 just treats the command as identical to a launch from the GUI. Dbfirs 21:20, 19 August 2016 (UTC)

@Dbfirs: @Vanamonde93: Very good. That's exactly the point. The one-off version I linked above, however, will/might actually print some debugging info to the terminal in the event an error condition like that Vandamonde described is encountered. Otherwise, the GUI operates identically. West.andrew.g (talk) 23:15, 19 August 2016 (UTC)

Oh, yes, I see. I can't get the latest version to go wrong, but perhaps someone else can? Dbfirs 23:41, 19 August 2016 (UTC)

Well, I followed the instructions above, and STiki launched its usual GUI. The very first edit I came across was this. I tried to revert AGF as I usually would, and the program dropped, as it has been: but no information of any sort appeared in my terminal window...not sure if that helps in any way. Vanamonde (talk) 09:25, 20 August 2016 (UTC)

Good faith revert is doomed

I have the latest version. If you click good faith, your session will expire. --Marvellous Spider-Man 10:33, 22 August 2016 (UTC)

There is a way. Don't click the good faith revert for first few edits. If it's good faith revert I clicked pass and reverted only vandalism. After dozen+ clicks, good faith revert works. --Marvellous Spider-Man 11:22, 22 August 2016 (UTC)

And after making the above comment, none of my Stiki edits are shown in My contributions in Wikipedia. I will close and try again. --Marvellous Spider-Man 11:28, 22 August 2016 (UTC)

The same is happening to Vanamonde93 (see above), and it happened to me with the old version of STiki, but not with the new version. We are all baffled by these crashes. Which version of Java are you using? (Mine is Version 8 update 101 (build 1.8.0_101-b13)) I've tried three different versions of Windows (Vista, 7 and 10) but I can't reproduce the error. Dbfirs 12:41, 22 August 2016 (UTC)

Yes, this is one of my major focuses right now. Please join the thread above. Unfortunately I don't have such an easy time reproducing the error on my own machine, so I am going to continue issuing one-off debugging versions to see if the effected folks can provide some feedback to me. Thanks, West.andrew.g (talk) 18:37, 22 August 2016 (UTC)

Strangely enough, the problem has disappeared for me in the last 12 hours or so: I'm reverting now with no issues. I'll let you know if the problem reappears. Vanamonde (talk) 05:27, 23 August 2016 (UTC)

Same here. Made 4-5 AGF edits and they all worked. I suspect this is a WMF issue and not a Stiki issue because the WMF servers do have their outages every now and then. --Rsrikanth05 (talk) 19:16, 23 August 2016 (UTC)

This is why my head hurts. West.andrew.g (talk) 03:13, 24 August 2016 (UTC)

The head-hurt you go through on our behalf is much appreciated :) Vanamonde (talk) 03:24, 24 August 2016 (UTC)

We're really really sorry for troubling you so much Andrew. We know the huge amount of effort that goes in in fixing issues with STiki. --Rsrikanth05 (talk) 08:40, 24 August 2016 (UTC)

Yes! Many, many thanks!!! Jacona (talk) 11:25, 24 August 2016 (UTC)

Edits doesn't show in my contribution

I make a series of vandalism reverts and then I check my contributions, which are still. This happen sometimes, not always. Marvellous Spider-Man 18:36, 24 August 2016 (UTC)

If I close the Stiki window, then after logging again, I can see that my Stiki edits are being saved in my contribution. --Marvellous Spider-Man 18:38, 24 August 2016 (UTC)

@Marvellous Spider-Man: This sounds like session loss similar to the AGF case being discussed above. Are you running the "2016_08_18" version? It was a special one-off version linked in an above section (and fairly well hidden, at that)... Are you running it from terminal? Thanks, West.andrew.g (talk) 19:30, 24 August 2016 (UTC)

Yes, I have downloaded the latest version few days ago, after reading the above posts. The latest version shows less errors then old version, but sometimes it doesn't work. I didn't understand Are you running it from terminal?. Marvellous Spider-Man 01:32, 25 August 2016 (UTC)

If same terminal means the same device, then; Yes, I am using the same device. Marvellous Spider-Man 03:22, 25 August 2016 (UTC)

No, Andrew means "are you running STiki from a DOS prompt (black screen in Windows) so that the exact error can be reported". You can get to a command prompt (DOS emulator) from Start --> Run --> CMD Dbfirs 06:42, 25 August 2016 (UTC)

Is there a chance anyone with good wikitext-foo would be willing to combine and collapse the two sections on this topic in a more organized fashion? Its becoming hard to follow. Ideally, maybe we could organize a nice little section that:, (a) more prominently links the latest one-off version, (b) collapses historical discussion, (c) has some quick notes on how to "run from terminal", and (d) contains a table with usernames and whether or not they are currently experiencing issues. Thanks, West.andrew.g (talk) 13:58, 25 August 2016 (UTC)

Option for make_comment_default automatically after reverting

I like adding additional info to the edit comment for good faith reverts, but I think the common case is putting in a new edit comment for every revert. I know make_comment_default is a private method of a different panel than gui_revert_panel, but is it possible to have an option to call make_comment_default automatically when reverting? Also, I'm not sure whether it's typical to change the edit comment, but the confusion over good faith edits in Wikipedia:Requests_for_adminship/Vanamonde93 maybe makes a good case for making modifying the edit comment a more prominent feature. – FenixFeather ^(talk)^(Contribs) 21:13, 29 August 2016 (UTC)

@FenixFeather: Greetings. I am a bit confused by this request. The "comments" panel has the capability to change the edit comment left with both "vandalism" and "AGF" reverts. It also has a "make default" button, that if pressed, will make whatever text is in the text box the default for that reversion type moving forward.

You had two requests: (1) "... modifying the edit comment a more prominent feature". It already has its own panel; I'm not sure what more I could do. (2) "... have an option to call make_comment_default automatically when reverting". Since you are already modifying the comment every time, is it that much work to click the button right below the box? I would imagine this workflow is the opposite of what most people do: Modify the comment in special circumstances, but generally want to return back to a generic comment moving forward, or at least as a generic starting point to craft the next customized message. If you are really modifying the comment at every AGF, what's the point of making the previous comment the default at all? Not trying to be defensive, I just don't understand your workflow. West.andrew.g (talk) 15:43, 1 September 2016 (UTC)

No worries. By prominence, I just mean maybe a mention in the documentation or a notification in the interface or something; in the RFA thread it seems like people are unaware that you can change the good faith revert edit comment, thus the confusion over why the candidate seemingly good faith reverted vandalism. As for my workflow, it is what you describe: using the default message there as a generic starting point to craft the next message (usually I just append a colon and put in what kind of GF revert I'm doing). If it resets after revert every time, it saves a bit of effort in manually removing the modification added to the edit comment. Not a huge deal, just a quality of life and optimizing for the common case sort of thing. – FenixFeather ^(talk)^(Contribs) 16:14, 1 September 2016 (UTC)

@FenixFeather: Ahh. It doesn't revert automatically back to the default generic? I just assumed that was the case based on how I thought I coded it. Thanks, West.andrew.g (talk) 22:40, 1 September 2016 (UTC)

Pass button

is making the program hang, as of the last 12 hours or so. Persists despite killing the process in task manager and restarting. Vanamonde (talk) 07:14, 4 September 2016 (UTC)

@Vanamonde93: I was away for the holiday weekend. I'll assume since no one else piled on here that the problem has resolved itself? It is unusual that this is specific to the "pass" button, as that action communicates exclusively with the STiki home server and not at all with the WMF (the other actions communicate with both). A database lock for the weekly STiki backup or the weekly production of the WP:5000 use/block a lot of resources on the STiki server, but those are for durations measured in minutes, not hours. Let me know if you observe the behavior again. West.andrew.g (talk) 15:18, 6 September 2016 (UTC)

I'm travelling right now, with poor internet: I'll let you know if this recurs once I get back to a good connection. Vanamonde (talk) 04:28, 7 September 2016 (UTC)

Name change on the leaderboard

Just to let you know I previously used to be known as Class455fan1, and had my name changed to "Class455" about three weeks ago. Class455 (talk) 10:25, 7 September 2016 (UTC)

Done: @Class455: Your 560 old contributions have been re-mapped to the new username. This will be reflected at tonight's leadboard re-gen. Thanks, West.andrew.g (talk) 18:25, 7 September 2016 (UTC)

Userpage notice

@West.andrew.g: I previously edited under the username The Pancake of Heaven!. Please re-map my contributions to my old account. Thanks! Wiki Pancake ^{✉ 📖} 13:10, 13 September 2016 (UTC)

Done -- @WikiPancake: 366 contributions re-mapped; will be reflected at tonight's leadboard re-generation. Thanks, West.andrew.g (talk) 13:40, 13 September 2016 (UTC)

Edits not saving etc

There seem to be several simultaneous bugs today: first, none of my edits are saving; second, the issue of STiki crashing during AGF reverts has popped up again, and today it's on every AGF revert. Finally, I got a spontaneous session drop message just after launching it the first time. Vanamonde (talk) 07:49, 13 September 2016 (UTC)

The rollback token does not appear to be working here as well. Ditto on AGF revert, but no session drops either. lavender|^{(formerly HMSSolent)}|lambast 01:32, 14 September 2016 (UTC)

Edit: Yesterday vandalism/AGF did not work at all; this morning, it seems to work after a restart, so likely AGF acting up again. lavender|^{(formerly HMSSolent)}|lambast 01:36, 14 September 2016 (UTC)

The funny thing in my case (this is still happening) is that vandalism reverts seem to work, but then do not show up in my contributions as they should. Vanamonde (talk) 05:29, 14 September 2016 (UTC)

Requesting permissions

Hello, I think I've found my calling on Wikipedia for the foreseeable future - anti-vandalism. As such I am requesting a mentor and permissions to use the tool. Thanks for the consideration. Justin15w (talk) 23:41, 16 September 2016 (UTC)

You've shown promise in the past few days fighting it. I'm unsure that's enough, moreover Andrew might not too. Then the two quickest ways for getting permission in your case would be: enrolling at the Academy which you should pass quickly or continuing your work till you reach say 500 edits (you're 300 approx right now) just doing such work and asking for rollback. Both these will surely grant you access. Good luck. Ugog Nizdast (talk) 14:22, 17 September 2016 (UTC)

I had similar thoughts. I spot checked your contributions, and they seemed good; but 89 article-space edits (which is what you had when I checked) is definitely on the low side. Just plug away at what you're doing, and you'll surely be given the tool sooner rather than later. Vanamonde (talk) 14:45, 17 September 2016 (UTC)

Understood. Thanks to both of you for the tips! Justin15w (talk) 16:07, 17 September 2016 (UTC)

@Justin15w: Yep, just a little more practice or WP:CVUA graduation and I'll speedily approve. Thanks, West.andrew.g (talk) 18:19, 17 September 2016 (UTC)

Question re: STiki and Twinkle

Stupid question - what functionality does STiki offer over Twinkle? Thanks in advance. Justin15w (talk) 23:23, 16 September 2016 (UTC)

I'm sure others can expand further, but here's the basic rundown. STiki finds and shows you edits that might be vandalism and then gives you the option to revert and warn (or mark as good faith, not vandalism, and so on). Twinkle just automates reverting and warning. It's on you to find vandalism to revert (through Special:Watchlist, Special:RecentChanges, etc.). clpo13_(talk) 23:27, 16 September 2016 (UTC)

Excellent, thank you sir. Justin15w (talk) 23:30, 16 September 2016 (UTC)

There is more info at Wikipedia:STiki#Comparison with other tools. The comparison was written with things like Huggle in mind, but most of the points apply to Twinkle too. Yaris678 (talk) 17:21, 23 September 2016 (UTC)

Request for approval

Hi there! I am a relatively new wiki user, and have been spending my time on here patrolling recent changes. However, manual patrolling has gotten quite repetitive and time consuming. A tool like this would allow me to streamline and overall help contribute to the wiki and patrol recent changes! Basically, i'm looking for approval and hope that someone can do that for me. Thanks, Netdoge (talk) 01:41, 27 September 2016 (UTC)

Netdoge: I took a look at your contributions, and based on a brief review you seem to have made a good start. However, you had 25 edits when I checked, and that is rather soon to ask for this tool. I would suggest plugging along for a little while longer. You should consider enrolling at the counter vandalism academy, which can help you refine your skills. Regards, Vanamonde (talk) 03:56, 27 September 2016 (UTC)

@Netdoge: Agreed on all fronts. WP:CVUA is a resource that can expedite our approval process. Thanks, West.andrew.g (talk) 19:02, 27 September 2016 (UTC)

Name change

I have previously used STiki under the username "Sunmist3". Can those contributions be remapped to "Sunmist"? Sunmist (talk) 03:19, 27 September 2016 (UTC)

Done -- @Sunmist: 272 contributions re-mapped; will be reflected at nightly report re-generation. Thanks, West.andrew.g (talk) 19:12, 27 September 2016 (UTC)

Edits not counting?

Every now and then (like tonight) I find my STiki edits aren't actually making any changes according to Contributions. Is there a reason? Thanks! Justin15w (talk) 22:07, 3 October 2016 (UTC)

@Justin15w: I'll point you to the recent archives for endless discussion of this problem. It's intermittent nature suggests the problem doesn't lie (solely) in the STiki code base. We've had difficulty reproducing the errors, and even spun-off special debugging editions of STiki to try to learn more. We haven't learned much. It also seems to affect a certain subset of users more frequently. We've had a good run recently of no one mentioning this error, although a few weeks ago, it was quite prevalent. ((shrugs shoulders)). The easiest thing here to blame something on the WMF server-side, but even if that's the case, I'd like for STiki to detect when we've encountered such a condition and mitigate the ill outcomes. West.andrew.g (talk) 19:07, 4 October 2016 (UTC)

Thank you kindly, Andrew, for the explanation. I'll read up in the archives. Justin15w (talk) 19:11, 4 October 2016 (UTC)