How slow is Discourse on your mobile?
-
WP8.1 Samsung Ativ S Neo on WiFi:
-
This was painful. Moto X, Chrome
I sent this page's url to mobile, it loaded pretty quick, but I almost clicked on the link multiple times because it took so long to respond. Then that page crawled. Took a screenshot, hit back, waited. Scrolling down fifty posts seemed daunting so I tried using the widget to go down. That was a mistake. I'd clicked it three times before it opened, then it closed right away. 2/5 would not Discourse again.
-
Galaxy Note 3 Neo (Android 4.4)
Internet Browser (Android standard)
Google Chrome
Toshiba Excite (Android 4.3)
Internet Browser (Android Standard)
Firefox Beta
Google Chrome
Tough I am really not sure what the test is supposed to solve.
Doing the Test multiple times resulted in very different results
Here my first try on Chrome with multiple tabs open (note that the result is 2k less)
So if you could enlighten me, @Sam, what numbers I am looking at and how they are to be interpreted (including the huge variance). That would really be appreciated.
-
No, mott's right. WinPhone 7 can't take screenshots unless put in developer mode. I've come across that myself before.
-
Note 4 and S 5 have been released in the last few months.
Are the Note 4 and the S5 basically the same phone, but just a difference in display?
I ask because my wife is up for a new phone and she likes my S5, but I wonder if I might like the Note 4 better, assuming it has the same hardware? With the difference in model numbers, it almost makes it seem like the Note is one generation behind the S-series in hardware.
-
A quick Google suggests the Note 4 is more powerful than the S5.
-
Notes tend to come out a few months after the S, and in some ways are test beds for the next S. For example, there's also a Note Edge this year, with some weird curved screen gimmick, and the rumours are that there will be an Edge version of the S6 as well..
Generally the Note N will be a bit higher spec than the S N+1, and also has the extra size and S Pen. Extra size also means a bigger battery. If the price difference isn't a factor and she likes a larger screen then go for the Note
-
If the price difference isn't a factor and she likes a larger screen then go for the Note
Well, she would be getting my ~3 month old S5 and I would be getting the Note 4.
The hard part will be talking her in to it, because she is going to see my phone as being "used", even though it does not have a scuff on it...
-
Same phone, default browser.
-
Jolla, Sailfish 1.0.8.21 - Firefox on Alien Dalvik
Same phone - Native Jolla Browser, based on Gecko.
-
How much do I get paid for testing your Belgium for you?
-
Hopefully with better performance?
-
iPod touch (4th generation) with iOS 5.1.1:
-
No, mott's right. WinPhone 7 can't take screenshots unless put in developer mode. I've come across that myself before.
Funny... I don't remember saying he was wrong...
-
Oh yeah, that's hilarious.
-
Then again, if you have a cool phone you can probably brag about it in this topic
I'm tempted to try it out on my HTC Hero when I get home.
Filed under: Different kind of cool
-
Mandatory Dinosaur References:
https://meta.discourse.org/t/what-will-computer-interfaces-look-like-in-the-future/18874/6
source: http://meta.discourse.orghttp://dilbert.com/dyn/str_strip/000000000/00000000/0000000/200000/30000/5000/100/235184/235184.strip.gif
source: http://dilbert.comFiled Under: 2009 ... that was probably a good year for something I am now too lazy to look up
-
my HTC Hero
Ahh, my first Android phone. It wasn't particularly snappy even for the time, being the same processor as the G1 but with the factory underclocking disabled. I had a kernel that allowed me to overclock to the glorious heights of 768 MHz
-
-
Ahh, my first Android phone. It wasn't particularly snappy even for the time, being the same processor as the G1 but with the factory underclocking disabled. I had a kernel that allowed me to overclock to the glorious heights of 768 MHz
I pretty much had to use a custom rom (with overclocking, natch) just to get the 2.1 update, as Telus (the Canadian telco that carried the phone delayed the update for months after the rest of the world had it.
Actually that reminds me, it's probably running some hacked-up ICS rom right now, so it'll be a treat to see how that affects the Discourse test.
-
I never found a full 2.2 rom. Everything I could find at the time indicated that there was no proper 2.2 kernel so everything that claimed to be 2.2 was 2.1 with cosmetic tweaks. The official 2.1 update for me required two separate download, wipe, flash cycles for some reason. No OTA or other user friendly options
-
It's been a million years, but I had no problem finding 2.2 (and later 2.3) roms on XDA, and they seemed legit. I'm not so sure about the ICS roms, but by the time those came out I had upgraded my phone so I didn't really care that much.
-
wow, what phone is this?
-
S3 Mini, that was one of the "Median" load while sat at the train station. There was another that was twice as long before the phone had cached the JS and CSS.
I might clone your repo and see where there are some easier JS optimizations. Because cracked.com with all it ads runs faster than Discourse.
-
Any help appreciated, keep in mind chrome, ember and us are all working on the android issue at the moment, the performance gap is huge though
-
The performance gap is huge because you are using a huge JS framework. It works great on desktop websites with good connections and fast processors ... but when you don't have that ....
-
My HTC Hero got the high score:
-
I ran it again on my Note 2 with the Android browser and it was 3100-3500 (I forget the exact number.) The Android browser wouldn't upload for some reason--I clicked the link but nothing happened. I was too lazy to reload the site in Chrome to post the picture.
-
Actually look at the iOS numbers here. It is very Android specific. Hearing good things about what the V8 Android team is doing in response to my blog post nudging. And the Ember team has a good benchmark suite now that we assisted with.
edit: and we're about to cut the number of posts / topics we display on Android in half for now, to keep speed up. This will mean a bit more loading as you scroll, but will also (as expected) reduce initial load time by half.
-
At the weekend I might have a proper look at how discourse is working on this phone and in chrome generally (iOS and Chrome is not better but that is due to how iOS and Chrome deals with memory leaks).
-
btw ... I am making lots of progress on making these cryptic backtraces usable
https://meta.discourse.org/t/production-source-maps-now-live/23077
-
http://i.imgur.com/wVMgXks.png
We're off to a promising start already!http://i.imgur.com/ZF27B2V.png
There we go.
Sony Xperia U, Android 4.04, Chrome 38.0.2125.114
-
Is there not some kind of framework that's a bit more lightweight you can serve to mobile devices?
It's not just Android (even if it is the worst offender) it's anything that isn't iOS based on the numbers here.
-
Oh, oh, I know the answer to this one!
It's using plain HTML and paging, right?
-
To be clear for anyone that hasn't caught on to the unit of ms: That screen loaded - to a "usable" state - in twenty-three seconds.
-
It's using plain HTML and paging, right?
You'd think, but I figured that idea would be a non-starter.
-
There is a no-js view, I think there is a chance of the browser detection javascript setting
window.location.href = window.location.href + "?no_js=1"
which would, ideally, remove the JS and the<noscript>
tag into a<noscropt>
tag.'noscropt' is probably a joke
referring toCnonection: close Connection: keep-alive
-
From memory you can't post in the non-JS view though?
-
Yeah, I figure a banner linking to
?no_js=0
would be enough, maybe?I clearly haven't
thought about thisdesigned this enough to actually do it.
-
To be clear for anyone that hasn't caught on to the unit of ms: That screen loaded - to a "usable" state - in twenty-three seconds
Then they should be running the tests on topics and not just front page. For me, topics quite frequently never enter a "usable state".
@sam, @codinghorror, can these tests be run on topics also? Anything with > 50 posts kills my phone.
-
@eviltrout is looking at this particular problem, we are going to cut down on posts that android renders which gives you faster topic render that are more jerky to scroll through.
medium term there are quite a few patches landing in ember that should help here and a bunch of stuff we can do.
-
You know more about your stack, but I am not sure this is the issue. When a topic does load and I get the the bottom, most recent post, it won't finish its rendering and show suggested topics. Just a spinner. Frequently it never even begins rendering the frame at all. Would that be mitigated by lowering the post count?
-
it's anything that isn't iOS based on the numbers here
Are we really taking Windows Phone seriously? And Blackberry is presumably using Blink/V8.
-
We should definitely get posts on the screen before suggested topics, hold back a run loop before showing them on android
-
My point was more "As this isn't just an Android-only issue why not focus on a change that improves it for all".
-
Beyond that, the only phone that produces numbers that aren't total rubbish is the latest gen iPhone. Even numbers for iPhone 5 are slower than I would consider acceptable. Load to a usable state should be under half a second. iPhone 5 is showing 3 times that.
-
Are we really taking Windows Phone seriously?
Are you seriously asking that question?
-
It is a well known fact <that i just made up>that Windows Phone is used entirely by furry enthusiasts. Jeff is a furry hater.
-
@Intercourse said:
It is a well known fact that Windows Phone is used entirely by furry enthusiasts. Jeff is a furry hater.
DISCRIMINATION!
-
Are furrys a protected class in the UK? :-P