On Wednesday 21 December 2005 23:46, Brion Vibber wrote:
Magnus, are you working with Tim on this or are you just writing code that's going to be thrown away next week?
Dear Brion,
I talked with Tim on this mailing list, where he said that stable versions and their templates need to be cached, and that he worked on some kind of patch.
I said OK, since you're working on that, I'll rest my extension.
A few days went by. As not much seemed to happen in this regard, I asked if someone was working on it. I received no reply.
So I went ahed and did was had been discussed on my own, since noone seems to do it.
And all I get as a reaction for, yet again, doing what everyone else just keeps talking about is yor above comment.
This seems to be a pattern. Last year, I realized that we'd need a review/validation feature, soon. So I went ahed and started implementing one. I checked it into CVS and kept updating and improving it. Release erly, release often, the open source mantra. I assumed that, this being open source, and having 60+ developers listed at sourceforge for MediaWiki, I would get some help in finishing it.
I didn't, not really. I wrote it up to working status basically on my own, working in comments from test site users.
I checked in the first version seventeen month ago. SEVENTEEN MONTH! And all the "help" I got was some vague uneasyness from you about it. At one point you said something I could work with, namely that you were concerned about it loading too much data from the database (or something along that line), which I promptly fixed.
Then, I think it was two weeks ago, you said that, after seventeen month of hot air, that the validation feature was unsalvagable. So I went ahead and rewrote it as a feature, which was one of your wishes some month ago. I have not heard anything about that from you either.
Don't take this as a personal attack. Maybe you're just overworked (you've done a lot of work to be sure). Maybe you're suffering from a mild case of Linus-Thorvalds-Syndrome, drowning in the responsibility for the code. Or maybe you just don't like me; but then, I'm not the only one to be treated like this, so it can't be that.
What *I* wish for is a process like this: * Someone writes a major new function/extension to be used on the Wikimedia sites * Someone in charge of these sites (Tim, Brion, whoever) says either ** "OK, that's great, we'll use it right away" OR ** "No, we'll never use that, so don't waste your time" OR ** "We would use it if you fix THIS and change THAT"
I once was an author at Nupedia. The approval process was so slow that few people ever wrote there, let alone got an article through the review process. A few of my articles made it afer three to four month. It was a drag, but my article was in the "official" list at the end.
But being largely ignored for SEVENTEEN MONTH, writing code and pushing for it to be used, and then being told "nope" would make paid programmers leave a company. I'm sure that, if not you youself, at least others on this mailing list might imagine what that kind of treatment can do for the motivation of a volunteer developer.
So, to come back to the original topic, this is *the* chance to answer a few of my questions: * Is /anyone/ actively working on a stable version feature? * If so, is it already more advanced than the code I put into CVS? * Is there some other reason this can't be done based on my code?
As I have said many times before, I'm not pushing these things because they are /my/ code. I'm pushing them because IMHO we need them, and AFAIK my code is the only one currently available that is up to the job. If I am in error on one of these two, please tell me so, and do it in clear, simple words, as I see myself unable to convert musings of vague discomfort into code changes.
Also, again, I have no problem if you think my code is crap. That might well be the case. *All I want* is that you tell me * what the problems are * if Wikimedia would use the code if I were to fix those problems, or if I shouldn't bother
I'll be on the 22C3, so maybe we could discuss these things in person.
Magnus
Hi Magnus,
I'll be on the 22C3, so maybe we could discuss these things in person.
Not everyone of us may come to 22C3, but everyone involved in site operation are discussing our changes in IRC channels constantly. In real life developers have proper meetings, we can't afford those, so online communication is preferred. Some features could be discussed using Skype as well.
Long email threads maybe can be tracked in future, but they do not help with proper cooperation and collaboration.
So far it was always absolutely efficient and easy to communicate any changes to other developers on #mediawiki or #wikimedia-tech.
Cheers, Domas
Domas Mituzas wrote:
Hi Magnus,
I'll be on the 22C3, so maybe we could discuss these things in person.
Not everyone of us may come to 22C3, but everyone involved in site operation are discussing our changes in IRC channels constantly. In real life developers have proper meetings, we can't afford those, so online communication is preferred. Some features could be discussed using Skype as well.
Long email threads maybe can be tracked in future, but they do not help with proper cooperation and collaboration.
So far it was always absolutely efficient and easy to communicate any changes to other developers on #mediawiki or #wikimedia-tech.
IRC is fine, but with different timezones it's hard to be uptodate. For example, tim & brion are chatting a lot early morning, there is a lot of activity on Sunday. Not everyone is able to be on IRC 24/24 to track all the activity (although we can still read logs). Also, Magnus is not on IRC, which doesnt help either.
Long emails are good, but wikitech-l is flooded with several issues. It can either be issues with the wikifarm or the mediawiki code.
I think we could use a dedicated mailing-list where only devs / sysadmin would be able to post (rest will be heavily moderated). Then when someone do something important, he can send an email to that list and will be sure everyone will know about it. The server admin log on wikitech is just about that, but for the farm only.
I would love to do developers meeting every 3 months or so, but with lot of us in europe and the core developers in australia and usa ... It's a bit hard (or we have to be really rich).
An example of "yet another tool" that we could use one day: http://www.dotproject.net/ (let you create projects with tasks, assign people to them etc..). Anyway, that might not be the best solution to solve a communication trouble :o)
cheers,
(thanks deadchip@bmpx for the dotproject.net link)
On 12/22/05, Ashar Voultoiz hashar@altern.org wrote:
I think we could use a dedicated mailing-list where only devs / sysadmin would be able to post (rest will be heavily moderated). Then when someone do something important, he can send an email to that list and will be sure everyone will know about it. The server admin log on wikitech is just about that, but for the farm only.
I tried setting up a developer-only IRC channel some time ago (#mediawiki-dev), it didn't work out at the time, for reasons that became obvious to me at a later point (the people who know what I'm talking about, well, know what I'm talking about:).
The reason I think it would be a good idea to have an open development channel that would be moderated (people would basically get +v if they contributed something more valuable than "how do I set my sidebar items" to the discussion) is that it would avoid a lot of the entropy we get from user support questions that drown out real development discussion in #mediawiki. I think it's commendable that people are interested in helping end users with their in-the-manual-you-could-have-found-it-with-two-minutes-of-searching-on-google-or-meta questions, I'm just not were interested in it, and I'm sure there are others that feel the same.
Anyway getting back to your suggestion, I sort-of-agree, we have private-l for well, private stuff:) and then we have the zoo at wikitech-l, one would think that there needs to be some middle ground, however I haven't seen it working out in practice, if we take #mediawiki and #wikimedia-tech for example a lot of the discussion that "should" be in one channel ends up in the other, and more often than not people just crosspost their support questions to all the forums rather than picking the appropriate one, if this is to be done it needs, as you suggested, to be moderated.
Not to butt too violently into this discussion, but...
I am pretty much ecstatic about everything that Mr. Manske has been working on in the last few weeks. Even if it's never accepted into the core codebase, as someone who's running his own wiki, just having some code that's marginally-official or at least run by this mailing list, to implement features I could really use, is fantastic.
Reviews? Great! Stable versions as an extension? Even better! The Tasks stuff is also tremendous..
I'm really looking forward to getting back to work after the New Year and getting a bunch of this stuff integrated. Maybe it's not the final solution - that's fine, I know how to use merge tools and if I have to I can write a converter to move my data over. I'm more concerned about reasonably solid code that people other than me are writing.
So from the perspective of an outside guy trying to run a site on MediaWiki - awesome job all around, I love seeing useful code being released often. Magnus' stuff is good, and it builds on a lot of cool work by many talented people.
I'm sure that there's plenty of activity relating to this discussion that I am not privy to, and I realize that the core mission for most people on this mailing list revolves around keeping the wikipedia family of sites happily up and running, but I don't want to let this discussion go by without tossing in my two cents as an outside user and say - I like seeing progress, I like the features I've seen go by recently, and I like MediaWiki as a whole.
I hope this whole conversation comes to a fruitful resolution, and that MW maintains its forward motion.
Thanks for reading. I'll go back to lurking now. :)
Ben Garney Torque Technologies Director GarageGames.Com, Inc.
Magnus Manske wrote:
On Wednesday 21 December 2005 23:46, Brion Vibber wrote:
Magnus, are you working with Tim on this or are you just writing code that's going to be thrown away next week?
At present, we have lots of people with CVS access, and it's up to each developer to decide when they commit to CVS HEAD. For example, if I don't hear back from Brion on the namespace changes soon, I'm going to give them a final look and merge them into HEAD. Brion usually _does_ react quickly to problematic changes there .. ;-)
Then there's patches by outside contributors without CVS access. I think these can be dealt with by allowing anyone with CVS access to handle them, since the assumption is that people with CVS access are reasonable. It may be desirable for some developers to take on a formal reviewer responsibility, so that people can contact them explicitly.
And then there's code that is in HEAD, but it is either an extension, or it is a site-wide option. In those cases, someone has to make an explicit decision that the code is ready for use on the Wikimedia sites. I believe your recent extensions, as well as the old Validate code, fall into that category.
In those cases, the general procedure tends to be "Wait for Brion" or "Bug someone on IRC", but it's not immediately clear what the responsibilities are.
It may make sense to let Brion give some people a formal "code activator" privilege. These people would be allowed to enable code on the live sites. Bugzilla could be used to keep track of who is going to review and possibly activate the code, if anyone. If there are no volunteers, and the code is generally deemed important and useful, the responsibility would fall back onto Brion.
This would let you know that someone is indeed going to review your code, and if they decline to integrate it at the present time, will give you feedback on what needs to be changed.
What do others think?
Erik
Erik Moeller wrote: <snip>
What do others think?
Erik
I haven't looked at the Stable extension so I cant comment on it. The task / revision extensions looks promising but need a bit more time before they are made available on live website.
Maybe it's time to release 1.6 with the current HEAD code, merge changes from wikidata branch and work on it (keeping live site under 1.6). We could then really concentrate on Stable / Tasks / Revision . I am not sure how brion decide when a new release should be created though.
Hopefully, when we switch to subversion, it will probably be a bit easier to play with branches.
cheers,
Ashar Voultoiz wrote:
Erik Moeller wrote:
<snip>
What do others think?
Erik
I haven't looked at the Stable extension so I cant comment on it. The task / revision extensions looks promising but need a bit more time before they are made available on live website.
See, that's part of the perception problem. You're right, these extensions are likely not ready for use as they are now.
But they don't need more *time*. They need more /work/! Someone has to do the work, either me (as I am probably most familiar with them), or someone else. I, for one, have done what I can think of right now. I'm willing to work, but I need feedback for what needs to be done. Preferrably from those who decide if and when to turn the extension on at the Wikimedia sites. Yes, work takes time, but we're not talking about wine here; having it age in the basement won't improve code.
We did have plenty of time, and we wasted much of it. Now we're running out of time, fast. If you have doubts about that, subscribe to the google news alert with keyword "wikipedia". I used to get a list of news each day how great wikipedia is. Today I got a list of half a dozen items telling me how flawed and outright dangerous wikipedia is. I've been getting these since the Steigentaler incident.
I am well aware of the press hype and the deconstruction of stars, but this will change public perception of Wikipedia. Along with that will come more and more "elements" that try to hit on the weak, making the situation even worse. I don't argue to rush blindly into action no matter what, but IMHO we (the technical people) really have to get our act together to offer some real options to the community. Which of them are put into use, for how long, and under which conditions, is another topic.
Magnus
Magnus Manske wrote: [snip]
Then, I think it was two weeks ago, you said that, after seventeen month of hot air, that the validation feature was unsalvagable. So I went ahead and rewrote it as a feature, which was one of your wishes some month ago. I have not heard anything about that from you either.
I haven't had a chance to go over that code yet; I've got to review that and Erik's wikidata stuff still. But I do feel a lot better about it being done as an extension, and I'd love to see it up and running soon.
What *I* wish for is a process like this:
- Someone writes a major new function/extension to be used on the
Wikimedia sites
- Someone in charge of these sites (Tim, Brion, whoever) says either
** "OK, that's great, we'll use it right away" OR ** "No, we'll never use that, so don't waste your time" OR ** "We would use it if you fix THIS and change THAT"
Yes, we need to be better about that. I know the most frustrating thing is waiting for a solid reaction, and I'm sorry I haven't been as good about that as I should.
Please call me on it if you see me being wishy-washy on things.
So, to come back to the original topic, this is *the* chance to answer a few of my questions:
- Is /anyone/ actively working on a stable version feature?
Tim has been working on it for about a week, and is about to commit into a branch.
- If so, is it already more advanced than the code I put into CVS?
We'll see when he commits his code, which should be Very Soon Now.
- Is there some other reason this can't be done based on my code?
Who knows. :D
I'll be on the 22C3, so maybe we could discuss these things in person.
Great, I'll see you there!
A few things:
* Those of us who are mostly chatting on IRC, we should remember to keep everybody posted on the mailing list when we start or finish something major.
* If actively working on code, popping into IRC is strongly recommended to get some immediate feedback. In particular, I'd like to ask that anyone making commits to CVS be in #mediawiki at the time to respond to questions or reactions that others may have.
* Those of us working on code for more than a day or two without it making it into CVS should show post work patches from time to time to show our progress. A live demo site is particularly helpful.
-- brion vibber (brion @ pobox.com)
Brion Vibber wrote:
Magnus Manske wrote: [snip]
Then, I think it was two weeks ago, you said that, after seventeen month of hot air, that the validation feature was unsalvagable. So I went ahead and rewrote it as a feature, which was one of your wishes some month ago. I have not heard anything about that from you either.
I haven't had a chance to go over that code yet; I've got to review that and Erik's wikidata stuff still. But I do feel a lot better about it being done as an extension, and I'd love to see it up and running soon.
Great!
What *I* wish for is a process like this:
- Someone writes a major new function/extension to be used on the
Wikimedia sites
- Someone in charge of these sites (Tim, Brion, whoever) says either
** "OK, that's great, we'll use it right away" OR ** "No, we'll never use that, so don't waste your time" OR ** "We would use it if you fix THIS and change THAT"
Yes, we need to be better about that. I know the most frustrating thing is waiting for a solid reaction, and I'm sorry I haven't been as good about that as I should.
OK, no hard feelings.
Please call me on it if you see me being wishy-washy on things.
I will! :-)
So, to come back to the original topic, this is *the* chance to answer a few of my questions:
- Is /anyone/ actively working on a stable version feature?
Tim has been working on it for about a week, and is about to commit into a branch.
- If so, is it already more advanced than the code I put into CVS?
We'll see when he commits his code, which should be Very Soon Now.
OK. As I said, I'm not trying to push my code here because /I wrote it/. Tim likely knows the MediaWiki code better than I do, so we'll probably end up using his solution. I have no problem with that at all, I'm just glad things get done.
- Is there some other reason this can't be done based on my code?
Who knows. :D
I'll be on the 22C3, so maybe we could discuss these things in person.
Great, I'll see you there!
Looking forward to it.
A few things:
- Those of us who are mostly chatting on IRC, we should remember to
keep everybody posted on the mailing list when we start or finish something major.
- If actively working on code, popping into IRC is strongly
recommended to get some immediate feedback. In particular, I'd like to ask that anyone making commits to CVS be in #mediawiki at the time to respond to questions or reactions that others may have.
Personally, I prefer mails; IRC always strikes me as ... hectic. I'll try to be there more often, though.
- Those of us working on code for more than a day or two without it
making it into CVS should show post work patches from time to time to show our progress. A live demo site is particularly helpful.
This is probably intended for Tim :-) as I have always tried to supply both. I'd like to remark that a *central* demo site might be more helpful. We do have the test site, but there should be an easy way to * update it from live CVS * turn extensions on and off, and alter the test database (add new tables for extensions etc.), for CVS developers
Maybe I should write some code for that. Then again, maybe not ;-)
Magnus
wikitech-l@lists.wikimedia.org