Jimbo's plan calls for everyone being able to rate, including not-logged-in users. This will be for test purposes only;
the ratings
from the test phase will be deleted again, once the
statistics people
got their hands on it and can hopefully tell us how to fine-tune the rating process.
The "performance hog myth" dates back to the very first
version. I have
worked on the problematic parts and IMHO they should be
good. I can't
be certain without real testing, though. However, I doubt it will degrade performance in the initial phase; if there's a
problem, it will
most likely show once there are at least a few hundred thousand ratings. There will be many write/delete queries on the table, which might lead to table locks; this might be countered with low priority queries, though.
Magnus
If testing this new feature does NOT cause the servers to slow down immediately, then there is no reason not to try it out. Someone had claimed an immediate TRIPLING of access time, but Magnus says "I doubt it will degrade performance in the initial phase", and I believe him.
Magnus is one of the three people whom we all trust the most at Wikipedia. If he says there's no risk in a live test of a new feature, then I say there's no risk.
Tim Starling, do you agree?
Ed Poor Former Developer
On 11/4/05, Poor, Edmund W Edmund.W.Poor@abc.com wrote:
If testing this new feature does NOT cause the servers to slow down immediately, then there is no reason not to try it out. Someone had claimed an immediate TRIPLING of access time, but Magnus says "I doubt it will degrade performance in the initial phase", and I believe him.
Magnus is one of the three people whom we all trust the most at Wikipedia. If he says there's no risk in a live test of a new feature, then I say there's no risk.
Tim Starling, do you agree?
Ed Poor Former Developer
If people are really worried we could:
1.)Try this on one of the smaller languaes
2.)try it around 3am on a monday
3.)I can't be the only one involved in a non wikimedia project. Give it to one of them to test. -- geni
--- geni geniice@gmail.com wrote:
If people are really worried we could:
1.)Try this on one of the smaller languaes
2.)try it around 3am on a monday
3.)I can't be the only one involved in a non wikimedia project. Give it to one of them to test.
Theres no issue of trust in the developers, but prudence is always prudent : I agree with Geni (in this case): simple.wikipedia.org would be a good place to do it, as would drive a little added interest in simple. (I dont know if thats a good idea or not).
SV
__________________________________ Yahoo! FareChase: Search multiple travel sites in one click. http://farechase.yahoo.com
Theres no issue of trust in the developers, but prudence is always prudent : I agree with Geni (in this case): simple.wikipedia.org would be a good place to do it, as would drive a little added interest in simple. (I dont know if thats a good idea or not).
I thought simple was going to get shut down.
Of course, it could still be used as a guinea pig for new features, but I didn't realise that there is still interest in reviving it.
On 11/4/05, Timwi timwi@gmx.net wrote:
Of course, it could still be used as a guinea pig for new features, but I didn't realise that there is still interest in reviving it.
It has it's uses. Allowing people to destress while still being productive for one.
-- geni
geni wrote:
On 11/4/05, Timwi timwi@gmx.net wrote:
Of course, it could still be used as a guinea pig for new features, but I didn't realise that there is still interest in reviving it.
It has it's uses. Allowing people to destress while still being productive for one.
If that is its only use, then doing so isn't productive. :-p
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
geni wrote:
On 11/4/05, Poor, Edmund W Edmund.W.Poor@abc.com wrote:
If testing this new feature does NOT cause the servers to slow down immediately, then there is no reason not to try it out. Someone had claimed an immediate TRIPLING of access time, but Magnus says "I doubt it will degrade performance in the initial phase", and I believe him.
Magnus is one of the three people whom we all trust the most at Wikipedia. If he says there's no risk in a live test of a new feature, then I say there's no risk.
Tim Starling, do you agree?
Ed Poor Former Developer
If people are really worried we could:
1.)Try this on one of the smaller languaes
2.)try it around 3am on a monday
3am in which timezone? UTC?
- -- Alphax | /"\ Encrypted Email Preferred | \ / ASCII Ribbon Campaign OpenPGP key ID: 0xF874C613 | X Against HTML email & vCards http://tinyurl.com/cc9up | / \
On 11/5/05, Alphax alphasigmax@gmail.com wrote:
3am in which timezone? UTC?
Yes from what I recall between 3 and 6 is pretty quite on that day.
-- geni
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
geni wrote:
On 11/5/05, Alphax alphasigmax@gmail.com wrote:
3am in which timezone? UTC?
Yes from what I recall between 3 and 6 is pretty quite on that day.
What about in other parts of the world? Not everone (especially people who use Wikipedia!) lives in Western Europe.
At 3AM UTC, the West Coast of the USA (UTC -8) is at 7PM, the East Coast of the USA (UTC -5) is at 10PM, with the rest of the US inbetween. Australia is between 11AM and 2PM depending on where you are and if it's daylight savings or not.
More helpful than speculating about "3am saturday morning sounds good" would be to check the server logs (assuming they exist) to get some ACTUAL DATA.
- -- Alphax | /"\ Encrypted Email Preferred | \ / ASCII Ribbon Campaign OpenPGP key ID: 0xF874C613 | X Against HTML email & vCards http://tinyurl.com/cc9up | / \
On 05/11/05, Alphax alphasigmax@gmail.com wrote:
More helpful than speculating about "3am saturday morning sounds good" would be to check the server logs (assuming they exist) to get some ACTUAL DATA.
http://noc.wikimedia.org/stats.php
Times appear to be given in UTC, judging by "Last updated" being only a couple of hours behind current UTC time - it looks like load is lowest around 0000 to 0600 UTC on Saturdays and Sundays, with Saturday generally being a quiet day.
So all things considered, 3am Saturday morning looks good ;-)
-- - Andrew Gray andrew.gray@dunelm.org.uk
Poor, Edmund W wrote:
If testing this new feature does NOT cause the servers to slow down immediately, then there is no reason not to try it out. Someone had claimed an immediate TRIPLING of access time, but Magnus says "I doubt it will degrade performance in the initial phase", and I believe him.
Magnus is one of the three people whom we all trust the most at Wikipedia. If he says there's no risk in a live test of a new feature, then I say there's no risk.
Thanks :-)
I'd like to add that *if* I'm wrong about this, someone flip a switch in LocalSettings to turn it off again. So, worst-case scenario is the servers getting slow for a few minutes.
Magnus
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
Magnus Manske wrote:
Poor, Edmund W wrote:
If testing this new feature does NOT cause the servers to slow down immediately, then there is no reason not to try it out. Someone had claimed an immediate TRIPLING of access time, but Magnus says "I doubt it will degrade performance in the initial phase", and I believe him.
Magnus is one of the three people whom we all trust the most at Wikipedia. If he says there's no risk in a live test of a new feature, then I say there's no risk.
Thanks :-)
I'd like to add that *if* I'm wrong about this, someone flip a switch in LocalSettings to turn it off again. So, worst-case scenario is the servers getting slow for a few minutes.
The servers just went slow, did you turn it on? :)
- -- Alphax | /"\ Encrypted Email Preferred | \ / ASCII Ribbon Campaign OpenPGP key ID: 0xF874C613 | X Against HTML email & vCards http://tinyurl.com/cc9up | / \
Alphax wrote:
Magnus Manske wrote:
I'd like to add that *if* I'm wrong about this, someone flip a
switch in
LocalSettings to turn it off again. So, worst-case scenario is the servers getting slow for a few minutes.
The servers just went slow, did you turn it on? :)
No, but I thought really hard about it :-)
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
Bryan Derksen wrote:
Magnus Manske wrote:
Alphax wrote:
The servers just went slow, did you turn it on? :)
No, but I thought really hard about it :-)
Aww. So once again, we go right up to the brink and then back off?
Throw ze svitch!
Seconded.
- -- Alphax | /"\ Encrypted Email Preferred | \ / ASCII Ribbon Campaign OpenPGP key ID: 0xF874C613 | X Against HTML email & vCards http://tinyurl.com/cc9up | / \
Alphax wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
Bryan Derksen wrote:
Magnus Manske wrote:
Alphax wrote:
The servers just went slow, did you turn it on? :)
No, but I thought really hard about it :-)
Aww. So once again, we go right up to the brink and then back off?
Throw ze svitch!
Seconded.
Go for it.
-- Neil
On 11/6/05, Neil Harris usenet@tonal.clara.co.uk wrote:
Alphax wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
Bryan Derksen wrote:
Magnus Manske wrote:
Alphax wrote:
The servers just went slow, did you turn it on? :)
No, but I thought really hard about it :-)
Aww. So once again, we go right up to the brink and then back off?
Throw ze svitch!
Seconded.
Go for it.
-- Neil
You know what you doing--- For Great Quality!
WikiEN-l mailing list WikiEN-l@Wikipedia.org To unsubscribe from this mailing list, visit: http://mail.wikipedia.org/mailman/listinfo/wikien-l
-----Original Message----- From: Stefan Sittler [mailto:kiaparowits@gmail.com] Sent: Monday, 7 November 2005 3:34 AM To: English Wikipedia Subject: Re: [WikiEN-l] Try out the rating scheme
No, but I thought really hard about it :-)
Aww. So once again, we go right up to the brink and then back off?
Throw ze svitch!
Seconded.
Go for it.
Be bold!
--Cyberjunkie
Cyberjunkie wrote:
Be bold!
Well, I guess just having a vote cascade on the mailing list isn't going to work. Hrmph. Anyone know who specifically is "in charge" of deciding what Wikimedia configuration to use? Or is the problem perhaps that there _isn't_ anyone specifically in charge? :)
Well, I guess just having a vote cascade on the mailing list isn't going to work. Hrmph. Anyone know who specifically is "in charge" of deciding what Wikimedia configuration to use? Or is the problem perhaps that there _isn't_ anyone specifically in charge? :)
Yes, of course that is the problem. There are several developers who could enable the feature, but all of them are now in fear that something will go Horribly Wrong™, so each individual one of them doesn't want to be held responsible in the end, so they don't do it and hope that someone else will do it and take the blame instead.
Which is kind of weird, because Brion and Tim have done much more "dangerous" things in the past, and every once in a while it does bring the servers down, but they always fix it swiftly and it's never a big deal.
This wouldn't be any different, had it not been for the initial FUD-laden discussion, which now leaves the developers fearful of social repercussions.
Timwi
Just find someone with serves access to a non mediawiki wiki and turn it on and see what happens. If all else fails the person who runs the other wiki project I work on would love to have this feature.
-- geni
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
geni wrote:
Just find someone with serves access to a non mediawiki wiki and turn it on and see what happens. If all else fails the person who runs the other wiki project I work on would love to have this feature.
I think you meant "non-wikimedia" - given that we're talking about features of Mediawiki...
Actually, given the number of customisations that have been carried out to our wikis, I'm not entirely sure that the results would be valid. We need the exact same conditions with only 1 variable changed (the presence of the validation feature) for the results to be valid in any way.
Given that Commons has traditionally been a place of innovation and trying out new features, why not start there?
- -- Alphax - http://en.wikipedia.org/wiki/User:Alphax Contributor to Wikipedia, the Free Encyclopedia "We make the internet not suck" - Jimbo Wales
On 11/9/05, Alphax alphasigmax@gmail.com wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
geni wrote:
Just find someone with serves access to a non mediawiki wiki and turn it on and see what happens. If all else fails the person who runs the other wiki project I work on would love to have this feature.
I think you meant "non-wikimedia" - given that we're talking about features of Mediawiki...
err yes.
Actually, given the number of customisations that have been carried out to our wikis, I'm not entirely sure that the results would be valid. We need the exact same conditions with only 1 variable changed (the presence of the validation feature) for the results to be valid in any way.
If we were looking at wether it would be a bulk server hog I think we could get away with that. Hardwear setups would be more of an issue but even limited data is better than none
Given that Commons has traditionally been a place of innovation and trying out new features, why not start there?
Because rateing the kind of material they have there doesn't make a hudge amount of sense. If we wanted to test it on a smaller wikimedia project then one of the non english wikipedias would be most logical. Which one would depend on the size of test you wanted to carry out.
-- geni
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
geni wrote:
On 11/9/05, Alphax alphasigmax@gmail.com wrote:
geni wrote:
<snip>
Actually, given the number of customisations that have been carried out to our wikis, I'm not entirely sure that the results would be valid. We need the exact same conditions with only 1 variable changed (the presence of the validation feature) for the results to be valid in any way.
If we were looking at wether it would be a bulk server hog I think we could get away with that. Hardwear setups would be more of an issue but even limited data is better than none
I was thinking more in terms of "given the number of extensions that have already been applied, will adding validation break them, be broken by them, or cause the whole thing to crash?" However I am reasonably certain that the devs have made sure this is not the case.
What I'm not convinced about is the complex server structure...
Given that Commons has traditionally been a place of innovation and trying out new features, why not start there?
Because rateing the kind of material they have there doesn't make a hudge amount of sense. If we wanted to test it on a smaller wikimedia project then one of the non english wikipedias would be most logical. Which one would depend on the size of test you wanted to carry out.
Actually that was a pretty stupid statement of me to make. Commons has done quite a few things wrt. multilingualism (cf. Babel) but in terms of actually getting content out there and available, de: has been the leader (cf. the DVD and such).
So: should we ask the de: community what they think? Try it out over there?
- -- Alphax - http://en.wikipedia.org/wiki/User:Alphax Contributor to Wikipedia, the Free Encyclopedia "We make the internet not suck" - Jimbo Wales
On 11/9/05, Alphax alphasigmax@gmail.com wrote:
Actually that was a pretty stupid statement of me to make. Commons has done quite a few things wrt. multilingualism (cf. Babel) but in terms of actually getting content out there and available, de: has been the leader (cf. the DVD and such).
So: should we ask the de: community what they think? Try it out over there?
If we are looking at limited roleout for a test .de is way too big something around 10,000 articles (not .nn) would probably be a good choice. .ko perhaps?
-- geni
On 09/11/05, geni geniice@gmail.com wrote:
On 11/9/05, Alphax alphasigmax@gmail.com wrote:
Actually that was a pretty stupid statement of me to make. Commons has done quite a few things wrt. multilingualism (cf. Babel) but in terms of actually getting content out there and available, de: has been the leader (cf. the DVD and such).
So: should we ask the de: community what they think? Try it out over there?
If we are looking at limited roleout for a test .de is way too big something around 10,000 articles (not .nn) would probably be a good choice. .ko perhaps?
Without meaning to sound Eurocentric - how easy is it for the .ko community to get back to the developers with problems? AIUI we're used to "experimental" work happening with .en, but not elsewhere...
Huh. Wait a second. We've already said, somewhere upthread, that the test implementation of the feature is solely to gather information on how it works. So why not enable it on meta and see what happens? Ratings will be meaningless, sure (though possibly amusing), but it lets people play with it, it lets us see if it works, and when whatever machine meta is on fails to explode into a ball of flaming debris then we can roll it out somewhere people might actually see it and rate articles "for real"...
-- - Andrew Gray andrew.gray@dunelm.org.uk
On 11/9/05, Andrew Gray shimgray@gmail.com wrote:
Without meaning to sound Eurocentric - how easy is it for the .ko community to get back to the developers with problems? AIUI we're used to "experimental" work happening with .en, but not elsewhere...
Hmm then simple english would be the other logical choice. It doesn't have the size I would like but there would be no language problems.
Othere options would be Wiktionary (bigger than I would like it to be) and wikibooks (which would be about the right size)
-- geni
Alphax wrote:
Actually, given the number of customisations that have been carried out to our wikis, I'm not entirely sure that the results would be valid. We need the exact same conditions with only 1 variable changed (the presence of the validation feature) for the results to be valid in any way.
Wikimedia has no test servers? All of them are production systems?
Andrew Venier wrote:
Wikimedia has no test servers? All of them are production systems?
We do have a test server, somewhere, which even ran an ancient version of the validation feature last time I looked.
The point seems to be to test the feature on a high-traffic site, which the test server is not.
Magnus
On 11/9/05, Andrew Venier avenier@venier.net wrote:
Wikimedia has no test servers? All of them are production systems?
The code has already been tested in such a fashion - the question is whether it can stand a production load.
Perhaps the true problem is that we do not have any method of simulating a production load adequately.
-Matt
Matt Brown wrote:
On 11/9/05, Andrew Venier avenier@venier.net wrote:
Wikimedia has no test servers? All of them are production systems?
The code has already been tested in such a fashion - the question is whether it can stand a production load.
Perhaps the true problem is that we do not have any method of simulating a production load adequately.
How did the FAA (I think it was them) simulate crashing a jet aircraft to determine if their new fuel was safer or not? By crashing one, of course.
On 11/8/05, geni geniice@gmail.com wrote:
Just find someone with serves access to a non mediawiki wiki and turn it on and see what happens. If all else fails the person who runs the other wiki project I work on would love to have this feature.
I would invite you to do it on WikiTree (wikitree.org), but we're still running 1.4 and my developer status (while promised) has not happened yet.
-- geni _______________________________________________ WikiEN-l mailing list WikiEN-l@Wikipedia.org To unsubscribe from this mailing list, visit: http://mail.wikipedia.org/mailman/listinfo/wikien-l
On 11/9/05, Wikiacc wikiacc@gmail.com wrote:
On 11/8/05, geni geniice@gmail.com wrote:
Just find someone with serves access to a non mediawiki wiki and turn it on and see what happens. If all else fails the person who runs the other wiki project I work on would love to have this feature.
I would invite you to do it on WikiTree (wikitree.org), but we're still running 1.4 and my developer status (while promised) has not happened yet.
I have contacted the owner of the development version of WikiTree (which runs 1.5) for possible use.
Wikiacc wrote:
I have contacted the owner of the development version of WikiTree (which runs 1.5) for possible use.
Any word on whether he's interested?
I'm hoping this isn't going to just slip away into forgotten-land again. Just earlier today I had yet another conversation with someone online in which they said "Wikipedia's nice and all, but you can never tell if someone's just inserted a load of nonsense into it." With this feature I'd finally have a snappy comeback to that. And having just dipped my toes into AfD for a few days, I'd love to see anything that might alleviate it. :)
Bryan Derksen wrote:
Wikiacc wrote:
I have contacted the owner of the development version of WikiTree (which runs 1.5) for possible use.
Any word on whether he's interested?
I'm hoping this isn't going to just slip away into forgotten-land again. Just earlier today I had yet another conversation with someone online in which they said "Wikipedia's nice and all, but you can never tell if someone's just inserted a load of nonsense into it." With this feature I'd finally have a snappy comeback to that. And having just dipped my toes into AfD for a few days, I'd love to see anything that might alleviate it. :)
I'm surprised you've still got any toes left.