Vector is still a miserable failure for mobile phone users.
Is there any timescale for this being fixed? At the very least, graceful degradation put into place?
Latest bug report (from a friend in a Facebook conversation):
"mazing. conservapedia still works on my mobile, wikipedia doesn't. "
"htc touch pro 2, windows mobile 6.1 pro, using internet explorer (because opera is a steaming pile of shit). not sure what version of IE...
yeah, wikipedia used to work marvelously, i've completely stopped using it since the update - i only... ever used it to look things up in the pub anyways, anywhere else i'd need better cites :P"
Please. Save our readers from having to use Conservapedia instead, just because Monobook works and Vector doesn't!
- d.
On Tue, Aug 17, 2010 at 7:17 PM, David Gerard dgerard@gmail.com wrote:
Vector is still a miserable failure for mobile phone users.
Is there any timescale for this being fixed? At the very least, graceful degradation put into place?
Latest bug report (from a friend in a Facebook conversation):
"mazing. conservapedia still works on my mobile, wikipedia doesn't. "
"htc touch pro 2, windows mobile 6.1 pro, using internet explorer (because opera is a steaming pile of shit). not sure what version of IE...
yeah, wikipedia used to work marvelously, i've completely stopped using it since the update - i only... ever used it to look things up in the pub anyways, anywhere else i'd need better cites :P"
Please. Save our readers from having to use Conservapedia instead, just because Monobook works and Vector doesn't!
Just switch to Discover :-) http://thenextweb.com/apps/2010/08/17/discover-by-cooliris-is-the-slickest-w...
Aren't most mobile browsers supposed to be redirected to mobile.wikipedia.org anyway?
BTW, we are slowly fixing Vector issues for obscure browsers. For example, I just recently fixed Vector for the PS3's browser. To fix these issues we need 2 things: A complete description of the environment and display behavior (preferably with a screenshot or even a photo of your cell phone screen), and a way to test it. For some mobile phones we can use emulators, but for others we have no easy way to test Vector on them.
Ryan Kaldari
On 8/17/10 1:20 PM, Magnus Manske wrote:
On Tue, Aug 17, 2010 at 7:17 PM, David Gerarddgerard@gmail.com wrote:
Vector is still a miserable failure for mobile phone users.
Is there any timescale for this being fixed? At the very least, graceful degradation put into place?
Latest bug report (from a friend in a Facebook conversation):
"mazing. conservapedia still works on my mobile, wikipedia doesn't."
"htc touch pro 2, windows mobile 6.1 pro, using internet explorer (because opera is a steaming pile of shit). not sure what version of IE...
yeah, wikipedia used to work marvelously, i've completely stopped using it since the update - i only... ever used it to look things up in the pub anyways, anywhere else i'd need better cites :P"
Please. Save our readers from having to use Conservapedia instead, just because Monobook works and Vector doesn't!
Just switch to Discover :-) http://thenextweb.com/apps/2010/08/17/discover-by-cooliris-is-the-slickest-w...
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
On 17 August 2010 21:39, Ryan Kaldari rkaldari@wikimedia.org wrote:
BTW, we are slowly fixing Vector issues for obscure browsers. For example, I just recently fixed Vector for the PS3's browser. To fix these issues we need 2 things: A complete description of the environment and display behavior (preferably with a screenshot or even a photo of your cell phone screen), and a way to test it. For some mobile phones we can use emulators, but for others we have no easy way to test Vector on them.
Doubt I can draft this user into such testing - as far as they're concerned, Wikipedia deliberately broke itself so therefore it's Wikipedia's stupidity. (And this is not an unreasonable description of how it happened.)
A suggestion: post to the techblog with precisely what you need from what classes of users. This can then be spread around the bogosphere and recruit all the help with browser issues you could want.
- d.
On 8/17/10 2:33 PM, David Gerard wrote:
...Wikipedia deliberately broke itself so therefore it's Wikipedia's stupidity. (And this is not an unreasonable description of how it happened.)
I'm not sure insulting the people who worked on Vector is the best way to motivated them to fix things. At the last company I worked for (which owns a couple of top 100 websites and has a staff of thousands), they didn't even care about their sites working right on iPhones. Considering the number of cell phones that Vector does work on, and the fact that it was implemented by a small handful of developers, I think they should be commended rather than insulted. But that's just me. (FYI, I had no involvement in Vector development.)
A suggestion: post to the techblog with precisely what you need from what classes of users. This can then be spread around the bogosphere and recruit all the help with browser issues you could want.
This is a useful suggestion and I'll see if it's doable.
Ryan Kaldari
On 18 August 2010 19:57, Ryan Kaldari rkaldari@wikimedia.org wrote:
I'm not sure insulting the people who worked on Vector is the best way to motivated them to fix things. At the last company I worked for (which owns a couple of top 100 websites and has a staff of thousands), they didn't even care about their sites working right on iPhones. Considering the number of cell phones that Vector does work on, and the fact that it was implemented by a small handful of developers, I think they should be commended rather than insulted. But that's just me. (FYI, I had no involvement in Vector development.)
It worked, then it broke, and Wikimedia didn't check beforehand and was slow to act on it. That's the sequence of events as it actually occurred. That doesn't look good to the end users, no matter how little the people who made the change like it.
- d.
I just looked and I could only find 2 bugs filed about Vector not working on mobile devices. The first was for it being broken in NetFront browsers. This was fixed less than a month after being reported (and within a week of Wikimedia getting a device to test on). The second was for Vector crashing some blackberry phones. This was traced to a bug on Blackberry's side (the same day the bug was reported). A work-around was implemented a week later. Further testing was done on eight different models of blackberries and across 3 different Firmware versions. It was determined that phones upgraded to the v5 Firmware no longer crashed. That doesn't sound like Wikimedia being "slow to act".
And for what it's worth, Vector was apparently tested on IEMobile. It worked in the tests, but of course it's impossible to test every version of every browser on every phone OS.
Thanks for sending the user agent string, that will be useful for filing a bug in Bugzilla.
Ryan Kaldari
On 8/18/10 12:10 PM, David Gerard wrote:
On 18 August 2010 19:57, Ryan Kaldarirkaldari@wikimedia.org wrote:
I'm not sure insulting the people who worked on Vector is the best way to motivated them to fix things. At the last company I worked for (which owns a couple of top 100 websites and has a staff of thousands), they didn't even care about their sites working right on iPhones. Considering the number of cell phones that Vector does work on, and the fact that it was implemented by a small handful of developers, I think they should be commended rather than insulted. But that's just me. (FYI, I had no involvement in Vector development.)
It worked, then it broke, and Wikimedia didn't check beforehand and was slow to act on it. That's the sequence of events as it actually occurred. That doesn't look good to the end users, no matter how little the people who made the change like it.
- d.
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
At 2010-08-18 21:50, Ryan Kaldari wrote:
I just looked and I could only find 2 bugs filed about Vector not working on mobile devices. The first was for it being broken in NetFront browsers. This was fixed less than a month after being reported (and within a week of Wikimedia getting a device to test on). The second was for Vector crashing some blackberry phones. This was traced to a bug on Blackberry's side (the same day the bug was reported). A work-around was implemented a week later. Further testing was done on eight different models of blackberries and across 3 different Firmware versions. It was determined that phones upgraded to the v5 Firmware no longer crashed. That doesn't sound like Wikimedia being "slow to act".
And for what it's worth, Vector was apparently tested on IEMobile. It worked in the tests, but of course it's impossible to test every version of every browser on every phone OS.
Thanks for sending the user agent string, that will be useful for filing a bug in Bugzilla.
BTW. PPK made some tests for mobile phones and their browsers: http://www.quirksmode.org/mobile/
He might shout a lot on devs ;-), but I think that the Foundation might convince him to make some tests in various phones (he has quite a lot of them). You would at least know for sure which must be switch to mobile version at all times and which should (might display some preference dialogue). Also dynamic switch to monobook for some phones might be nice.
Just some thoughts.
Regards, Nux.
I think you guys are experiencing the java-script load issues, Due to our amount of javascript that is loaded some browsers (eg: IE6 on desktop) can't handle it and refuses to load and errors out, This <s>should</s> will be fixed when the work on the resource loader branch is finished and merged.
"K. Peachey" p858snake@yahoo.com.au writes:
I think you guys are experiencing the java-script load issues
Why is the mobile redirect left to Javascript? Wouldn't it be better for all concerned if the redirect happened before any PHP was loaded?
Wouldn't it be better for those older phones with little memory if the redirect happened in PHP?
Why is JS even involved in this?
http://svn.wikimedia.org/viewvc/mediawiki/trunk/extensions/WikimediaMobile/M...
Mark.
On Thu, Aug 19, 2010 at 11:52 AM, 49342) mah@everybody.org wrote:
Why is the mobile redirect left to Javascript? Wouldn't it be better for all concerned if the redirect happened before any PHP was loaded?
Wouldn't it be better for those older phones with little memory if the redirect happened in PHP?
Forget memory -- the extra network and processing time involved in fetching an extra complete HTML+CSS+JS version of the page, then throwing it away and fetching another one, is wasteful and annoying even on highly capable phones!
However, we can't just use a User-Agent check in PHP for this, since 90-something percent of the time the PHP scripts are not being executed: the Squid caches respond to most web requests directly.
So what would be required would be some filtering in the caches to check for particular User-Agents or other settings and send them the redirect directly, or send them through to PHP for possible redirection. (Assuming there's no problem with downstream caching, which I think should usually be ok the way we have things marked -- as long as the redirect responses are marked as private-cache or uncacheable.)
A JavaScript hack was quicker to put in place than the cache-level logic, but it was only ever meant as a stopgap.
-- brion vibber (brion @ pobox.com)
Brion Vibber wrote:
However, we can't just use a User-Agent check in PHP for this, since 90-something percent of the time the PHP scripts are not being executed: the Squid caches respond to most web requests directly.
So what would be required would be some filtering in the caches to check for particular User-Agents or other settings and send them the redirect directly, or send them through to PHP for possible redirection. (Assuming there's no problem with downstream caching, which I think should usually be ok the way we have things marked -- as long as the redirect responses are marked as private-cache or uncacheable.)
A JavaScript hack was quicker to put in place than the cache-level logic, but it was only ever meant as a stopgap.
-- brion vibber (brion @ pobox.com)
I was precisely thinking on this the other day. The javascript is just doing a regex on the user agent, and squid is perfectly capable of doing that.
yup, its seem fairly doable and the issue has been raised to our ops team. we're waiting to hear back about what's necessary in order to make it work.
--tomasz
On Aug 20, 2010, at 3:59 PM, Platonides wrote:
Brion Vibber wrote:
However, we can't just use a User-Agent check in PHP for this, since 90-something percent of the time the PHP scripts are not being executed: the Squid caches respond to most web requests directly.
So what would be required would be some filtering in the caches to check for particular User-Agents or other settings and send them the redirect directly, or send them through to PHP for possible redirection. (Assuming there's no problem with downstream caching, which I think should usually be ok the way we have things marked -- as long as the redirect responses are marked as private-cache or uncacheable.)
A JavaScript hack was quicker to put in place than the cache-level logic, but it was only ever meant as a stopgap.
-- brion vibber (brion @ pobox.com)
I was precisely thinking on this the other day. The javascript is just doing a regex on the user agent, and squid is perfectly capable of doing that.
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
I got the browser string!
Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) Vodafone/1.0/HTC_Touch_Pro2/1.14.161.6 (50207)
The user adds:
"that's it. :) the mobile site works. but i am not going to bother using it. *effort*. it doesn't show up in google results and i have epic cba."
Not sure what the bit about Google results means ...
- d.
On Wed, Aug 18, 2010 at 3:29 PM, David Gerard dgerard@gmail.com wrote:
"that's it. :) the mobile site works. but i am not going to bother using it. *effort*. it doesn't show up in google results and i have epic cba."
Not sure what the bit about Google results means ...
I think they mean that when they Google something, it gives them a link to the article on en.wikipedia.org, not the mobile site. You should probably tell them not to worry and that they'll automatically be redirected to the article on the mobile version (m.en.wikipedia.org).
A bug has been set up for the IEMobile issue: https://bugzilla.wikimedia.org/show_bug.cgi?id=24862
Tomasz was unable to reproduce the bug using the same browser version, but different phone. Perhaps K. Peachey is right and it depends on how much memory the phone is allocating to the browser.
Ryan Kaldari
On 8/18/10 12:29 PM, David Gerard wrote:
I got the browser string!
Mozilla/4.0 (compatible; MSIE 6.0; Windows CE; IEMobile 8.12; MSIEMobile 6.0) Vodafone/1.0/HTC_Touch_Pro2/1.14.161.6 (50207)
The user adds:
"that's it. :) the mobile site works. but i am not going to bother using it. *effort*. it doesn't show up in google results and i have epic cba."
Not sure what the bit about Google results means ...
- d.
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
On 19 August 2010 02:19, Ryan Kaldari rkaldari@wikimedia.org wrote:
Tomasz was unable to reproduce the bug using the same browser version, but different phone. Perhaps K. Peachey is right and it depends on how much memory the phone is allocating to the browser.
That appears to have been the problem with older BlackBerrys, i.e. the browser just ran out of heap.
- d.
wikitech-l@lists.wikimedia.org