meta: see web almanac for any insights about manifest usage

marcoscaceres commented 3 years ago

It might be good to look over: https://almanac.httparchive.org/en/2020/pwa#web-app-manifests

And see if there is anything to be gleamed from it that might influence things in the spec.

I'm cc'ing @hemanth as they may provide additional insights or just for their interest.

hemanth commented 3 years ago

Thanks for tagging.

@tunetheweb might have the latest big quires of the 2021 insights.

tunetheweb commented 3 years ago

We're in the middle of writing those right now for the 2021 edition. Follow along here: https://github.com/HTTPArchive/almanac.httparchive.org/issues/2153

However we did add a custom metric to the monthly HTTP Archive crawl to make it easier to query this data every month, so we can see what the make up of manifests are for the top 7-8million websites that HTTP Archive tracks monthly. Let me know if there's something specific you're looking for and can maybe help you query it.

aarongustafson commented 2 years ago

We (Edge) just overhauled our own manifest crawler and I have a dashboard showing usage, errors, store readiness, and more. Happy to share the details I find. It would also be great to correlate the data with what y’all are seeing too. Currently, we’re tracking a little over 135k manifests. Some of the highlights:

A little over 5% were invalid/un-parsable (mostly—over 97%—due to issues in the page’s HTML)
More than a few sites are struggling with icon sizes, typically providing incomplete values, using the multiplication sign or asterisk instead of x or X, including commas between multiple size entries, or including the string "apple-touch-icon-" before the size value.
Nearly 6% of manifests do not have a name.
Only 30% define a start_url
63% are "display: standalone"
Only 68% have a theme color
64% define a background color
Most (89%) do not define orientation
PNGs are the most popular icon format and most manifest define either 2 (43%) or 6 (~24%) icons
Only 6% define a language and 3.75% define a text direction

Per #399, I have been continuing to track query string usage and have not found many UUIDs embedded in the start_url… yet. The only ones I’ve identified are:

https://www.geekmagazine.com.br/?tracking=5eb941b082681 https://www.aadvantageeshopping.com/?source=mn|AA|ALL|mn|NA|bookmark|na||mobile|00000000 https://shopping.mileageplus.com/?source=mn|UA|ALL|mn|NA|bookmark|na||mobile|00000000 /?tracelog=51606102_9527_7259_7770&from=desktop /?ucid=HPS-1027 /?source=NV_PWA_HL_HP&agentcode=00399206&utm_medium=WebPush&utm_source=NotifyVisitor /?mc=xQZpn7mAMXLZ

I need to investigate these to see if they are using a truly unique identifier though. Most have "utm_source=webapp" or similar. The closest I’ve seen to a tracking value so far is the occasional "utm_campaign=X" or the language or similar. One accidentally (?!) posted an API Key in the start url.

I’ve also got data on newer features too (share_target, protocol_handlers, etc.). I’m planning to write-up a "State of the Manifest" some time in the next quarter.

marcoscaceres commented 2 years ago

Amazing stuff @aarongustafson! Thanks for gathering those stats.

Thanks also @tunetheweb! Will follow along.

tunetheweb commented 2 years ago

FYI we've run the stats for this years Web Almanac PWA chapter and the data is available here: https://docs.google.com/spreadsheets/d/16AkIdDBBkCR5Kgb7kyfYvnNLQBu23Vsh7MUSFHW9RtA/#gid=398503119

Let me know if you spot anything that looks odd and can investigate further.

christianliebel commented 1 year ago

The 2022 data can be found here: https://docs.google.com/spreadsheets/d/1PbzjhN--jU9MGuWobw5L9EsmlVzI9tlbCe3_NKA7giU/edit#gid=2077755325

kenchris commented 1 month ago

Closing due to inactivity

w3c / manifest

meta: see web almanac for any insights about manifest usage #987