Hi everyone,
Those of you who were at the Mobile quarterly review heard me mention Facebook Lite, an app that's designed especially for the developing world.
Notably, their app has a lot of optimisations which make it good for users in developing world:
- It's only 252kB, good for limited data plans. - It supports down to Android 2.2, good for older devices. - It's data-efficient, good for 2G connections and for people on limited data plans.
From a development perspective, some advantages are:
- You no longer have to support older versions of Android in your main app. - You can tailor the performance of the lite app to the older devices so it's faster. - You can tailor the features of the lite app to the developing market.
So obviously there are a lot of advantages for our users if we do this. And, selfishly, I can't stress enough how much dropping Android 2.3 from our current app would speed up development. As an example, almost all of the edge cases with lead images occurred on 2.3 devices, and they required quite a lot of investigation and hacking to fix them up. Obviously we've not dropped 2.3 so far because it's a very strategically important part of our user base, which I'm sure Carolynne can attest to!
I'd say that we should put some serious thought into whether we'd prefer to have a Wikipedia Lite app for the developing world, rather than our current "one app to rule them all".
Comments? Questions?
Dan
Hi Dan,
First, thanks for asking for input before making this decision.
Second, it would be good to get feedback directly from the 2.x users before making a major change, although I am not knowledgable about how major this change would be for the UX. Any ideas on how to get feedback at reasonable scale? This might require some creative outreach to our low-bandwith and/or non-English-speaking users, and I'm wondering if Fundraising, Legal, CA or the Wikipedia Zero team have ideas.
Cheers,
Pine
On Jan 30, 2015 9:45 PM, "Dan Garry" dgarry@wikimedia.org wrote:
Hi everyone,
Those of you who were at the Mobile quarterly review heard me mention Facebook Lite, an app that's designed especially for the developing world.
Notably, their app has a lot of optimisations which make it good for users in developing world:
- It's only 252kB, good for limited data plans.
- It supports down to Android 2.2, good for older devices.
- It's data-efficient, good for 2G connections and for people on
limited data plans.
From a development perspective, some advantages are:
- You no longer have to support older versions of Android in your main
app.
- You can tailor the performance of the lite app to the older devices
so it's faster.
- You can tailor the features of the lite app to the developing market.
So obviously there are a lot of advantages for our users if we do this. And, selfishly, I can't stress enough how much dropping Android 2.3 from our current app would speed up development. As an example, almost all of the edge cases with lead images occurred on 2.3 devices, and they required quite a lot of investigation and hacking to fix them up. Obviously we've not dropped 2.3 so far because it's a very strategically important part of our user base, which I'm sure Carolynne can attest to!
I'd say that we should put some serious thought into whether we'd prefer to have a Wikipedia Lite app for the developing world, rather than our current "one app to rule them all".
Comments? Questions?
Dan
-- Dan Garry Associate Product Manager, Mobile Apps Wikimedia Foundation
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
That sounds like it may be the way to go!
For iOS:
Probably no time for a lite version this quarter, but maybe the current version could be made lighter?
It could actually be a relatively simple thing to do. In fact, I just did a quick experiment:
Our current iOS app weighs in at *4.38 MB*.
By simply removing the splash images the app binary size drops to *2.37 MB*.
iOS 8 has some fancy new abilities to present non-images as splash screens, so I say we do this for iOS 8, drop the splash images for older devices, and pay very close attention to the change in binary size that results from any external libraries we use.
We can also migrate a couple more images used by the iOS app to glyphs in our font - which is an easy process with the scripts I wrote a while back. This will save a bit more space. We could also do a couple spikes to see what other low-hanging fruit there is for trimming the binary size.
I think we could get to under 2 MB without breaking a sweat, or even the need for a separate version.
On Fri, Jan 30, 2015 at 9:45 PM, Dan Garry dgarry@wikimedia.org wrote:
Hi everyone,
Those of you who were at the Mobile quarterly review heard me mention Facebook Lite, an app that's designed especially for the developing world.
Notably, their app has a lot of optimisations which make it good for users in developing world:
- It's only 252kB, good for limited data plans.
- It supports down to Android 2.2, good for older devices.
- It's data-efficient, good for 2G connections and for people on
limited data plans.
From a development perspective, some advantages are:
- You no longer have to support older versions of Android in your main
app.
- You can tailor the performance of the lite app to the older devices
so it's faster.
- You can tailor the features of the lite app to the developing market.
So obviously there are a lot of advantages for our users if we do this. And, selfishly, I can't stress enough how much dropping Android 2.3 from our current app would speed up development. As an example, almost all of the edge cases with lead images occurred on 2.3 devices, and they required quite a lot of investigation and hacking to fix them up. Obviously we've not dropped 2.3 so far because it's a very strategically important part of our user base, which I'm sure Carolynne can attest to!
I'd say that we should put some serious thought into whether we'd prefer to have a Wikipedia Lite app for the developing world, rather than our current "one app to rule them all".
Comments? Questions?
Dan
-- Dan Garry Associate Product Manager, Mobile Apps Wikimedia Foundation
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
(Oh, the splash images I'm talking about on the iOS app are only shown at startup and only for the brief second it takes the app to load. The reason they take up so much space is older versions of iOS made you include one version for your image for each screen dimension and density - that is, one sized for 3.5 inch phones, one for 3.5 retina, iPad & iPad retina, iPad mini & retina etc...)
On Fri, Jan 30, 2015 at 11:37 PM, Monte Hurd mhurd@wikimedia.org wrote:
That sounds like it may be the way to go!
For iOS:
Probably no time for a lite version this quarter, but maybe the current version could be made lighter?
It could actually be a relatively simple thing to do. In fact, I just did a quick experiment:
Our current iOS app weighs in at *4.38 MB*.
By simply removing the splash images the app binary size drops to *2.37 MB*.
iOS 8 has some fancy new abilities to present non-images as splash screens, so I say we do this for iOS 8, drop the splash images for older devices, and pay very close attention to the change in binary size that results from any external libraries we use.
We can also migrate a couple more images used by the iOS app to glyphs in our font - which is an easy process with the scripts I wrote a while back. This will save a bit more space. We could also do a couple spikes to see what other low-hanging fruit there is for trimming the binary size.
I think we could get to under 2 MB without breaking a sweat, or even the need for a separate version.
On Fri, Jan 30, 2015 at 9:45 PM, Dan Garry dgarry@wikimedia.org wrote:
Hi everyone,
Those of you who were at the Mobile quarterly review heard me mention Facebook Lite, an app that's designed especially for the developing world.
Notably, their app has a lot of optimisations which make it good for users in developing world:
- It's only 252kB, good for limited data plans.
- It supports down to Android 2.2, good for older devices.
- It's data-efficient, good for 2G connections and for people on
limited data plans.
From a development perspective, some advantages are:
- You no longer have to support older versions of Android in your
main app.
- You can tailor the performance of the lite app to the older devices
so it's faster.
- You can tailor the features of the lite app to the developing
market.
So obviously there are a lot of advantages for our users if we do this. And, selfishly, I can't stress enough how much dropping Android 2.3 from our current app would speed up development. As an example, almost all of the edge cases with lead images occurred on 2.3 devices, and they required quite a lot of investigation and hacking to fix them up. Obviously we've not dropped 2.3 so far because it's a very strategically important part of our user base, which I'm sure Carolynne can attest to!
I'd say that we should put some serious thought into whether we'd prefer to have a Wikipedia Lite app for the developing world, rather than our current "one app to rule them all".
Comments? Questions?
Dan
-- Dan Garry Associate Product Manager, Mobile Apps Wikimedia Foundation
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
Love the idea, and I agree with everything Monte said. We might also need to drop some 3rd party libs to go super-ultra light, depending on their size. Quick inspection shows the following:
- AFNetworking: ~500 KB - hpple: 41 KB
We'll need to be careful adding too many other frameworks to the light version, but we can use a separate target for it which doesn't link to 3rd party code.
More importantly, we'll also need to thoroughly analyze CPU usage (primarily animations) and network efficiency—cache misses and extra round trips will kill the experience.
Excited to talk about this next quarter!
Brian
On Fri, Jan 30, 2015 at 11:45 PM, Monte Hurd mhurd@wikimedia.org wrote:
(Oh, the splash images I'm talking about on the iOS app are only shown at startup and only for the brief second it takes the app to load. The reason they take up so much space is older versions of iOS made you include one version for your image for each screen dimension and density - that is, one sized for 3.5 inch phones, one for 3.5 retina, iPad & iPad retina, iPad mini & retina etc...)
On Fri, Jan 30, 2015 at 11:37 PM, Monte Hurd mhurd@wikimedia.org wrote:
That sounds like it may be the way to go!
For iOS:
Probably no time for a lite version this quarter, but maybe the current version could be made lighter?
It could actually be a relatively simple thing to do. In fact, I just did a quick experiment:
Our current iOS app weighs in at *4.38 MB*.
By simply removing the splash images the app binary size drops to *2.37 MB*.
iOS 8 has some fancy new abilities to present non-images as splash screens, so I say we do this for iOS 8, drop the splash images for older devices, and pay very close attention to the change in binary size that results from any external libraries we use.
We can also migrate a couple more images used by the iOS app to glyphs in our font - which is an easy process with the scripts I wrote a while back. This will save a bit more space. We could also do a couple spikes to see what other low-hanging fruit there is for trimming the binary size.
I think we could get to under 2 MB without breaking a sweat, or even the need for a separate version.
On Fri, Jan 30, 2015 at 9:45 PM, Dan Garry dgarry@wikimedia.org wrote:
Hi everyone,
Those of you who were at the Mobile quarterly review heard me mention Facebook Lite, an app that's designed especially for the developing world.
Notably, their app has a lot of optimisations which make it good for users in developing world:
- It's only 252kB, good for limited data plans.
- It supports down to Android 2.2, good for older devices.
- It's data-efficient, good for 2G connections and for people on
limited data plans.
From a development perspective, some advantages are:
- You no longer have to support older versions of Android in your
main app.
- You can tailor the performance of the lite app to the older
devices so it's faster.
- You can tailor the features of the lite app to the developing
market.
So obviously there are a lot of advantages for our users if we do this. And, selfishly, I can't stress enough how much dropping Android 2.3 from our current app would speed up development. As an example, almost all of the edge cases with lead images occurred on 2.3 devices, and they required quite a lot of investigation and hacking to fix them up. Obviously we've not dropped 2.3 so far because it's a very strategically important part of our user base, which I'm sure Carolynne can attest to!
I'd say that we should put some serious thought into whether we'd prefer to have a Wikipedia Lite app for the developing world, rather than our current "one app to rule them all".
Comments? Questions?
Dan
-- Dan Garry Associate Product Manager, Mobile Apps Wikimedia Foundation
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
I also like the idea, but I think Android is a much better initial target, as it is much more common in the low bandwidth market from what I gathered. On Jan 31, 2015 8:52 AM, "Brian Gerstle" bgerstle@wikimedia.org wrote:
Love the idea, and I agree with everything Monte said. We might also need to drop some 3rd party libs to go super-ultra light, depending on their size. Quick inspection shows the following:
- AFNetworking: ~500 KB
- hpple: 41 KB
We'll need to be careful adding too many other frameworks to the light version, but we can use a separate target for it which doesn't link to 3rd party code.
More importantly, we'll also need to thoroughly analyze CPU usage (primarily animations) and network efficiency—cache misses and extra round trips will kill the experience.
Excited to talk about this next quarter!
Brian
On Fri, Jan 30, 2015 at 11:45 PM, Monte Hurd mhurd@wikimedia.org wrote:
(Oh, the splash images I'm talking about on the iOS app are only shown at startup and only for the brief second it takes the app to load. The reason they take up so much space is older versions of iOS made you include one version for your image for each screen dimension and density - that is, one sized for 3.5 inch phones, one for 3.5 retina, iPad & iPad retina, iPad mini & retina etc...)
On Fri, Jan 30, 2015 at 11:37 PM, Monte Hurd mhurd@wikimedia.org wrote:
That sounds like it may be the way to go!
For iOS:
Probably no time for a lite version this quarter, but maybe the current version could be made lighter?
It could actually be a relatively simple thing to do. In fact, I just did a quick experiment:
Our current iOS app weighs in at *4.38 MB*.
By simply removing the splash images the app binary size drops to *2.37 MB*.
iOS 8 has some fancy new abilities to present non-images as splash screens, so I say we do this for iOS 8, drop the splash images for older devices, and pay very close attention to the change in binary size that results from any external libraries we use.
We can also migrate a couple more images used by the iOS app to glyphs in our font - which is an easy process with the scripts I wrote a while back. This will save a bit more space. We could also do a couple spikes to see what other low-hanging fruit there is for trimming the binary size.
I think we could get to under 2 MB without breaking a sweat, or even the need for a separate version.
On Fri, Jan 30, 2015 at 9:45 PM, Dan Garry dgarry@wikimedia.org wrote:
Hi everyone,
Those of you who were at the Mobile quarterly review heard me mention Facebook Lite, an app that's designed especially for the developing world.
Notably, their app has a lot of optimisations which make it good for users in developing world:
- It's only 252kB, good for limited data plans.
- It supports down to Android 2.2, good for older devices.
- It's data-efficient, good for 2G connections and for people on
limited data plans.
From a development perspective, some advantages are:
- You no longer have to support older versions of Android in your
main app.
- You can tailor the performance of the lite app to the older
devices so it's faster.
- You can tailor the features of the lite app to the developing
market.
So obviously there are a lot of advantages for our users if we do this. And, selfishly, I can't stress enough how much dropping Android 2.3 from our current app would speed up development. As an example, almost all of the edge cases with lead images occurred on 2.3 devices, and they required quite a lot of investigation and hacking to fix them up. Obviously we've not dropped 2.3 so far because it's a very strategically important part of our user base, which I'm sure Carolynne can attest to!
I'd say that we should put some serious thought into whether we'd prefer to have a Wikipedia Lite app for the developing world, rather than our current "one app to rule them all".
Comments? Questions?
Dan
-- Dan Garry Associate Product Manager, Mobile Apps Wikimedia Foundation
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
-- EN Wikipedia user page: https://en.wikipedia.org/wiki/User:Brian.gerstle IRC: bgerstle
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
For sure. Always fun to think about optimising :)
On Jan 31, 2015, at 12:26 AM, Yuri Astrakhan yastrakhan@wikimedia.org wrote:
I also like the idea, but I think Android is a much better initial target, as it is much more common in the low bandwidth market from what I gathered.
On Jan 31, 2015 8:52 AM, "Brian Gerstle" bgerstle@wikimedia.org wrote: Love the idea, and I agree with everything Monte said. We might also need to drop some 3rd party libs to go super-ultra light, depending on their size. Quick inspection shows the following:
AFNetworking: ~500 KB hpple: 41 KB We'll need to be careful adding too many other frameworks to the light version, but we can use a separate target for it which doesn't link to 3rd party code.
More importantly, we'll also need to thoroughly analyze CPU usage (primarily animations) and network efficiency—cache misses and extra round trips will kill the experience.
Excited to talk about this next quarter!
Brian
On Fri, Jan 30, 2015 at 11:45 PM, Monte Hurd mhurd@wikimedia.org wrote: (Oh, the splash images I'm talking about on the iOS app are only shown at startup and only for the brief second it takes the app to load. The reason they take up so much space is older versions of iOS made you include one version for your image for each screen dimension and density - that is, one sized for 3.5 inch phones, one for 3.5 retina, iPad & iPad retina, iPad mini & retina etc...)
On Fri, Jan 30, 2015 at 11:37 PM, Monte Hurd mhurd@wikimedia.org wrote: That sounds like it may be the way to go!
For iOS:
Probably no time for a lite version this quarter, but maybe the current version could be made lighter?
It could actually be a relatively simple thing to do. In fact, I just did a quick experiment:
Our current iOS app weighs in at 4.38 MB.
By simply removing the splash images the app binary size drops to 2.37 MB.
iOS 8 has some fancy new abilities to present non-images as splash screens, so I say we do this for iOS 8, drop the splash images for older devices, and pay very close attention to the change in binary size that results from any external libraries we use.
We can also migrate a couple more images used by the iOS app to glyphs in our font - which is an easy process with the scripts I wrote a while back. This will save a bit more space. We could also do a couple spikes to see what other low-hanging fruit there is for trimming the binary size.
I think we could get to under 2 MB without breaking a sweat, or even the need for a separate version.
On Fri, Jan 30, 2015 at 9:45 PM, Dan Garry dgarry@wikimedia.org wrote: Hi everyone,
Those of you who were at the Mobile quarterly review heard me mention Facebook Lite, an app that's designed especially for the developing world.
Notably, their app has a lot of optimisations which make it good for users in developing world: It's only 252kB, good for limited data plans. It supports down to Android 2.2, good for older devices. It's data-efficient, good for 2G connections and for people on limited data plans. From a development perspective, some advantages are: You no longer have to support older versions of Android in your main app. You can tailor the performance of the lite app to the older devices so it's faster. You can tailor the features of the lite app to the developing market. So obviously there are a lot of advantages for our users if we do this. And, selfishly, I can't stress enough how much dropping Android 2.3 from our current app would speed up development. As an example, almost all of the edge cases with lead images occurred on 2.3 devices, and they required quite a lot of investigation and hacking to fix them up. Obviously we've not dropped 2.3 so far because it's a very strategically important part of our user base, which I'm sure Carolynne can attest to!
I'd say that we should put some serious thought into whether we'd prefer to have a Wikipedia Lite app for the developing world, rather than our current "one app to rule them all".
Comments? Questions?
Dan
-- Dan Garry Associate Product Manager, Mobile Apps Wikimedia Foundation
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
-- EN Wikipedia user page: https://en.wikipedia.org/wiki/User:Brian.gerstle IRC: bgerstle
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l