They do not "presume" anything. They state clearly that the old version of the library was causing performance problems. They also state that 32k is noticeable on slow connections, which it absolutely is. In fact, on unstable connections, it can mean the difference between the site loading and not loading. Reliability is critical for government services. They also mention that the work improved the codebase and the developers' experience with it, which is a criminally underrated means of improving efficiency and therefore reducing cost.
All in all, this sounds like fantastic work relative to what we normally expect from government end user software. We should encourage this kind of thing, not jump to the hand-wavy "my tax dollars!" complaint, which approaches meaninglessness in its commonality.
They do presume this is a problem when they don't quantify it. At the end of the article it is made clear there is no quantification going on:
> We’ll provide more detail about exactly what that means in the next blog post.
The whole premise of the article is conjecture until we see the performance numbers presented. The claim is that removing a one-time load 32kB library is useful to page load times and user experience for low end users. It has ZERO specifications or performance numbers cited.
Besides which, the net effect of more data on a page is that it takes a little longer to load the page. For someone who has a slow phone, they are conditioned to this through use of other sites (or even clicking on the housing link at the top of the UK Gov site and waiting for a 360kB image to load). What is the advantage of speeding it up for them for a single site they may or may not use?
Will we also take it upon ourselves to justify anything that makes them click on the wrong link and have to wait to go back? Unlikely.
It’s not the loading that’s a major problem, it’s the runtime parse cost on low-end Android devices. Everyone has to use government services, so you don’t get to ignore cheap devices.
(Also, luckily the UK government spent zero dollars on fixing this.)
It's an accessibility problem. GOV.uk is the official government website that contains key information that must be made available to every UK resident. If shaving off 32K of JavaScript makes the website easier to load for people with extremely bad hardware, then it's worth it.
gov.uk is an absolute delight to use exactly because of this kind of focus on the details and valuing of simplicity. I can’t think of another website that is more useable, and that is super important for a site that needs to work for a very diverse user base who have no choice but to use it.
I think you’d also be surprised how little this is likely to have cost compared to benefit to the tens of millions of users of these systems. GDS run a tight ship.
All in all, this sounds like fantastic work relative to what we normally expect from government end user software. We should encourage this kind of thing, not jump to the hand-wavy "my tax dollars!" complaint, which approaches meaninglessness in its commonality.