With a feature where arbitrary aspects "update" from under you in ways you could never expect and ensure beforehand are locked down in your favor in the contract/agreement.
For example, a widget might have a compile bug where the CSS rules to explicitly lay out component size only get included in media queries requested with devices using such low PPI that you never catch it in testing because your devices are not icky enough. This "bug" might then be "fixed" in an "improvement release" down the track that ships the explicit sizing to everybody... including your page layouts, which are now broken.
Or maybe you could ship a calendar control that includes a fuzzy date parser, and then remove the date parser "because it was confusing too many users". All emails from power users pointing out lost productivity because of their inability to specify dates like "2 1/2 days ago" or "-2h", or having "2018" map to today's date last year and "yesterday" also include the current time, are quietly sent to /dev/null. The calendar control is updated to use Material Design, two days of which are spent building a test jig to autoreload on 18 devices at once using remote debugging, as part of a subfeature sprint to ensure a certain cubic-bezier transition animation (which 4 people have collectively watched 537 times) "flows" correctly with the right... twang. A further day is spent turning an Arduino into a virtual bluetooth keyboard to ping the 18th device in the aforementioned testing rig; it stubbornly refused to not go to sleep, even when plugged in.
A decentralized bidding infrastructure could be implemented that uses Bitcoin to allow users to vote on which UI designs they like the most by sending money to different wallets. Much ado is had about how to rebuild a Merkle tree from first principles on top of the blockchain in order to implement the wallet tracking needed to power the aggregate collection needed to power the feature. After A/B testing (and some shady network sniffing) finds the UI library is only being used by people with 1Gbps (fiber/5G) Internet connections on devices with 8-core CPUs, it is decided to simply download the entire Bitcoin blockchain into the Web browser of every consumer of the UI toolkit to perform the analytics about which style to load on startup. No testing is performed on smartphones, and in the standup meeting for the sprint in question nobody remembers to point out that 8-core ARM CPUs are not the same as 8-core i9s. The stubborn team lead has sufficient karma to retain the feature even after the stunned bugreports flow in.
exikyut|6 years ago
For example, a widget might have a compile bug where the CSS rules to explicitly lay out component size only get included in media queries requested with devices using such low PPI that you never catch it in testing because your devices are not icky enough. This "bug" might then be "fixed" in an "improvement release" down the track that ships the explicit sizing to everybody... including your page layouts, which are now broken.
Or maybe you could ship a calendar control that includes a fuzzy date parser, and then remove the date parser "because it was confusing too many users". All emails from power users pointing out lost productivity because of their inability to specify dates like "2 1/2 days ago" or "-2h", or having "2018" map to today's date last year and "yesterday" also include the current time, are quietly sent to /dev/null. The calendar control is updated to use Material Design, two days of which are spent building a test jig to autoreload on 18 devices at once using remote debugging, as part of a subfeature sprint to ensure a certain cubic-bezier transition animation (which 4 people have collectively watched 537 times) "flows" correctly with the right... twang. A further day is spent turning an Arduino into a virtual bluetooth keyboard to ping the 18th device in the aforementioned testing rig; it stubbornly refused to not go to sleep, even when plugged in.
A decentralized bidding infrastructure could be implemented that uses Bitcoin to allow users to vote on which UI designs they like the most by sending money to different wallets. Much ado is had about how to rebuild a Merkle tree from first principles on top of the blockchain in order to implement the wallet tracking needed to power the aggregate collection needed to power the feature. After A/B testing (and some shady network sniffing) finds the UI library is only being used by people with 1Gbps (fiber/5G) Internet connections on devices with 8-core CPUs, it is decided to simply download the entire Bitcoin blockchain into the Web browser of every consumer of the UI toolkit to perform the analytics about which style to load on startup. No testing is performed on smartphones, and in the standup meeting for the sprint in question nobody remembers to point out that 8-core ARM CPUs are not the same as 8-core i9s. The stubborn team lead has sufficient karma to retain the feature even after the stunned bugreports flow in.