I've been checking this out and I have a few thoughts I'd like to share (basically, what Bryan said):
- This feels like google's attempt to be again the hub for a big part of the web, to gather more information and data for it's ad business. True, anybody could spin up a amp cache system, but they will have the fastest/best/biggest one. - This is just using a subset of HTML, and severely limiting any interactivity, only whitelisted pixel tracking and amp approved ad services allowed. It is even worse that the 2002 web. If you really wanted to, you can subset what you send to mobile browsers and get the same benefits (provided you use a really good CDN). - Check the FAQ out, ample support for user tracking, ad serving and paywalls. Supporting this is supporting another closed for-profit web with Google at the front of all users.
On the surface, AMP *seems* extremely heavy handed. It’s restricting for
developers and puts publishers in a squeeze over the technology they can use — right now, many are heavily leveraging Javascript for tracking and other experiences on their sites. On the other hand, AMP offers far more flexibility than both Apple and Facebook do, is hugely beneficial for users and faster load times should lead to keeping even more of them. AMP might finally push the Web in a much more positive direction — one that’s faster, with less cumbersome JavaScript everywhere.
*http://thenextweb.com/insider/2015/10/07/googles-plan-to-speed-up-the-mobile... http://thenextweb.com/insider/2015/10/07/googles-plan-to-speed-up-the-mobile-web-burn-it-down/*
I really don't see how this could be a good idea for us in it's current form.
That said, it would be great to keep in mind the ideas and apply them to our platform to guide us on how to serve our users better.
On Fri, Oct 9, 2015 at 2:39 AM, Toby Negrin tnegrin@wikimedia.org wrote:
Hi Luis --
I honestly don't see a lot of difference between Google, Twitter and Facebook, since they are all ad supported entities with a fiscal responsibility to track their users and sell the data. Apple's a bit different on the surface since they have a different business model. I agree that these are bad for the internet but so are incredibly slow web pages that make apps essentially required for a good experience.
On the analytics, this would probably not include their use of our content in the knowledge graph or elsewhere and also might be troublesome for those who prefer google not to track their reading.
Bryan's ticket is a good embarkation point for thinking about supporting new clients; Reading is also planning some Reading infrastructure work for the summit which could relate[1]
-Toby
[1] https://phabricator.wikimedia.org/T114542
On Thu, Oct 8, 2015 at 2:02 PM, Luis Villa lvilla@wikimedia.org wrote:
Toby -
I'm generally 1000% on-board with slow follower for anything user-facing. The only reason I might make an exception here is because the competitors you mention are all pretty awful for the web generally, and this has uptake already from Google and Twitter. (Two isn't great, but two + slim opportunity for growth is way better than the guaranteed never-greater-than-1 we'll see from FB's option.)
The other reason this intrigues me is that if Google builds in some analytics, it might give us a better sense of their current usages for us than we currently have. Not much, obviously, but at least something. (Remember that in this scenario - direct access from Google properties - they already have all that information, the only question is whether it gets shared with us so that we can do something useful with it.)
That said, if implementing it is non-trivial, it doesn't make sense to spend a huge number of cycles to fast-follow. Hopefully some of the improvements Bryan mentions will make it easier in the future - it certainly doesn't look like we're in a world where the number of front ends is going to get smaller any time soon.
Luis
On Thu, Oct 8, 2015 at 1:25 PM, Toby Negrin tnegrin@wikimedia.org wrote:
Thanks Bryan and Pine.
My feeling is that there are many many new interfaces and form factors emerging right now and we should be cautious about adoption. For example Facebook's instant articles, apple news and even snapchat have similar offerings the AMP.
They all seem to be focusing on article speed in a landscape where most pages are larded up with a variety of trackers, ads and other scripts (which we don't have, although we have our own challenges on performance) with the ultimate goal of owning the delivery platform.
I'm nervous about picking winners in such a landscape although I'm excited about prototypes like things like the Apple Watch app that came out of the Lyon hackathon. I feel like a slow follower model where we see which solution if any becomes widely used is more appropriate for us.
-Toby
On Thursday, October 8, 2015, Pine W wiki.pine@gmail.com wrote:
Hi Bryan,
Ah, I was thinking of the 2 different mobile web editing experiences (not 2 different apps) for Android depending on form factor. My understanding is that tablets have VE enabled on mobile web now (I have yet to try it) while phones do not have VE enabled on mobile web yet.
Pine On Oct 8, 2015 12:56 PM, "Bryan Davis" bd808@wikimedia.org wrote:
On Thu, Oct 8, 2015 at 1:32 PM, Pine W wiki.pine@gmail.com wrote:
We currently have at least 6 channels, I believe:
- Desktop Web
- Mobile Web
- Android phone
- Android tablet
I don't think that we have separate native apps for the phone and tablet form factors.
- IPhone
- Legacy Android phone
I'm no expert on mobile developmemt, but perhaps WMF could
experiment with
Google's idea on just one channel to start?
AMP would only be appropriate for the mobile web channel from the list above. Implementing it would be a matter of placing some sort of translating proxy between MediaWiki and the requesting user agent that downgraded the HTML produced by MediaWiki to AMP's restricted HTML dialect. That sort of translation might be possible in MobileFrontend but it would likely be accomplished much more easily using some other tech stack that had good support for manipulation of HTML like a node.js service. It might be an interesting prototype project for a volunteer to experiment with a frontend app that consumed the RESTBase provided Parsoid HTML (e.g. https://en.wikipedia.org/api/rest_v1/page/html/NOFX) and spit out AMP compliant documents.
The only other option really to produce alternate HTML from MediaWiki would require swapping out the existing layer that translates an article's wikitext to HTML with a version that spoke AMP instead. That would be related to https://phabricator.wikimedia.org/T114194.
Bryan
Bryan Davis Wikimedia Foundation bd808@wikimedia.org [[m:User:BDavis_(WMF)]] Sr Software Engineer Boise, ID USA irc: bd808 v:415.839.6885 x6855
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l
-- Luis Villa Sr. Director of Community Engagement Wikimedia Foundation *Working towards a world in which every single human being can freely share in the sum of all knowledge.*
Mobile-l mailing list Mobile-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mobile-l