DEV Community

Discussion on: The weirdly obscure art of Streamed HTML

Collapse
peerreynders profile image
peerreynders • Edited on

I believe this is why Kroger.com used a SPA in the first place — if disparate teams’ APIs can’t be trusted, at least they won’t affect other teams’ code.

I think this is an aspect that assisted SPA adoption in general. Given that at the time MPAs were relatively slow to respond (server frameworks probably played a part as well), a ready-to-go (smallish) package of JS could have been shipped to the client and started doing some useful work.

Now when discussing web performance SPA proponents often counter that tuning with respect to FCP and TTI only effects initial page load — missing the point that if page loads are universally fast you may not need that SPA - unless you're going offline-first.

Back in 2013 Ilya Grigorik talked about "Breaking the 1000ms Time to Glass Mobile Barrier" and here we are 9 years later where multi-second load times on mobile are not unusual.

While now dated (2013: during 4G adoption) he describes the complexity of sending a packet over the cellular network which increases the latency for mobile compared to regular networks (sometimes I'm surprised anything works).

Picard facepalm - all that for a single TCP packet ...

He also points out that when it comes to web page load times reducing latency matters more than increasing bandwidth (More bandwidth doesn't matter (much)).

Dust, a template language that seems to have died twice.

Patrick Steele-Idem (2014): Marko versus Dust

"Marko was developed to address some of the limitations of Dust as well as add a way to associate client-side code with specific templates to rein in jQuery, provide some much-needed structure, and provide reusable widgets. Version 1.0 was released in May 2014." from eBay's UI Framework Marko Adds Optimized Reactivity Model

I think one reason is the inconsistent name

Basically if you didn't geek out over HTTP servers and the HTTP spec you probably didn't know or think about "chunked transfer-encoding" (in connection with HTML responses). And since about 2016 online searches would funnel you to the Streams API.

Collapse
tigt profile image
Taylor Hunt Author

missing the point that if page loads are universally fast you may not need that SPA - unless you're going offline-first.

The rest of this series will essentially be illustrating “are you sure you need that SPA?”

Offline-first is usually the purview of SPAs, or at best an app-shell model bolted onto a content-heavy website. However, I was able to do some mad science with Marko… its compiled-for-server representation only has to write to an outgoing stream. You probably see where this is going. (More on that later.)


Ilya’s work definitely inspired me. Subsequent MPA interactions luckily don’t have to do the full mobile connection TCP warmup if you use Connection: keepalive, and if you have analytics pinging your origin every 30 seconds anyway, that socket rarely gets torn down. We’ll see some of my measurements around that later.


Great point about the Streams API. Maybe we need an entirely new term altogether.