The problem I had was that the entire experience became really slow when I had it all working together as one "response". Sending back all that HTML for an entire form with some large select lists as well as the response of results became noticeable laggy when there was more than a half dozen results.
I ended up switching to Alpine Ajax (https://alpine-ajax.js.org/) and pulled the form out of the response and just used a local x-data on it to track the state. This greatly reduced the HTML I needed to send back to just the list of results. I did make the form a bit more complicated but the experience felt a lot snappier. Both versions just synced the form state from the URL and kept the initial render as full HTML from the server.
I found that Alpine + Alpine Ajax was SMALLER than HTMX 4 even though (in my opinion) it offers a lot more features in a more approachable and intuitive way if you need to do interactivity and don’t want to trigger a request just to toggle some classes or attributes. Of course you can use both together (I started down that road) but you are mixing worlds and making the bundle size bigger at the same time.
I still like HTMX and will probably use it again. I just found that with an interactive experience like a product listing page, where the HTML response was quite large/expensive to fetch, it wasn’t the best choice for that.
https://github.com/dsego/ssr-playground/blob/main/src/server...
I think dynamically fetching results (ala infinite scroll) would have been more complicated to build. I would have had to add an intersection detection (HTMX 4 has that) to trigger the form/request again, but with an ever increasing chunk offset slice. I also need a way to track that state as well. Considering I had one big HTML response, I don’t think I could have done that efficiently without splitting it anyway. Even if I could "pluck" the results I needed from the HTML response, the penalty of waiting for the response was the factor.
If you have any links to some demos where there is something like that, I would appreciate it.
All that being said, the count of items actually isn't the most relevant part. The most relevant part is the amount of HTML per single result. If you have a sophisticated card component (a title, a description, images, a table of information inside of it or complicated layout) just loading six cards could end up being hundreds of HTML elements.
So even if you lazy loaded elements in chunks, those chunks are still gonna take a while to fetch given the size of the HTML. You also break CTRL-F-ing since those results don’t exist in the DOM until you trigger their fetching code. Same problem as classic pagination.
I would have to completely refactor my entire experience regardless because if I want to lazily load the cards/results gradually, I would have to split up the layout to make it more efficient.
So my approach to shrink the HTML response was to break it up the "live" section to just the right half.
With HTMX you can do partial rendering and live updates via server-sent events. This gets you most of the way to the "client side" feel where things load dynamically based on user actions.
The only properly dynamic SPA-like feature in a forum I can think of is a WYSIWYG editor, but that you can build as a web component. Maybe a flexible highlight and quote system would be a bit difficult in pure HTMX (think of the comment functionality in Medium posts). So you'd want to build a few things in client side JS. But the main experience could very well be built with HTMX.
After initial page load, a service worker could intercept all the HTMX app's AJAX requests, but it could only work with what's already cached. Using a lot of tooling like workbox the developer could pre-cache the entire app. (Would that mean no dynamic responses?)
Alternatively, implement the entire HTMX app in JS, build the app once for the NodeJS/Deno backend and once for the front-end. After page load, the frontend could take over.
Or something like that. It's a bit convoluted as PWA solutions often are.
Have you found nice workarounds for that?
If you'd like to find out how I build Mobile-first apps with Go, HTML and HTMX hit me up.
Example: Render a scrollable list and you need to change contents of the li items mid scroll - it resets the scroll position to the top because of how it replaced the entire HTML on render. There are workarounds you can do, DOM hacks, but they aren’t ideal.
React is designed around that kind of thing. They do some funny magic under the hood to support seamless DOM reconciliation and things like prop drilling of entire components on the prop of another, and so-on. This gives them a more granular way to manipulate the DOM without total page re-rendering.
Next.js SSR works because it’s designed around React. So they’re giving you all the SSR stuff you want but you still get client-side React for rich interaction design and other SPA niceties.
So if someone releases an SSR-like thing like htmx it feels like something is missing.
Maybe htmx needs an opinionated framework that gets more into high performance rendering and state management.
Or maybe they hone in on the “basic web pages” market. Who knows, maybe those make a comeback somewhere. PDF gets replaced with standalone HTML or something, and people use HTMX, or maybe it’s used in a dev environment for HTML email design in some future where email can have events (who knows). Couple of ideas.
Right now it feel incomplete or has nowhere for me to plug it in. Love the concepts though, HTML/JS should have been like that.
This issue is reporting the same kind of thing - text selection is lost on re-render
Hono + WebComponents + HTMX + serverless is the backend for my apps now.
although i wonder if LLM changes things
i am interested in moving away from React but its also what LLMs know infinitely more of than HTMX
I think this tells all about the level of competence of these devs. That’s literally unmaintainable. So instead of fixing the problem and have a nice api + interface + ssr, they patch it up with a yet worse approach in htmx.
> there's a lot of forum software out there that still does the old way of rendering as much as possible on the server and using some JavaScript on client to improve its interactivity here and there. And people are happy with that
Yeah, server returns class xxx and jquery needs to target that. Once is out of sync silent bugs will appear. And will be many, not a few. No way to check at compile time.
WHY WILL NOT MY FORUM RUN
MISAGO MONGO DJANGO BUN
PERFECT CONFIGURATION
MISAGO MONGO DJANGO BUN
IF YOU ARE THE CHOSEN ONE
MISAGO MONGO DJANGO BUN
REWRITE MY APP AND MAKE IT RUN!
But on serious note, htmx is basically a solution in the search of a problem. It is the new hype.
Or rather, a solution that overlooks 2 decades of learnings. Yes, for a small set of projects htmx is okay, but even then, where htmx is ideal, static is king, and once static is not good enough, htmx sooner or later starts to feel like the XAML and BPEL soap.
The fundamental problem is that it is pretending to be a declarative language while entirely imperative.
But Misago sure does look good.
I'd extend this to "web apps that revolve around forms, views, filters and pagination" when it comes to "Ajax" stuff (yes, I'm old).
The part of htmx that's relevant to these topics being discussed here, I once wrote an inferior clone of, some years ago.
It was "isomorphic", tailored to our PHP-based CMS, and could handle filters, search forms, and pagination.
Suitable for forums definitely, and current PHP-based forum software vendors that are still on the market follow the same paradigm.
I'd say that "Endless Scroll" is a good example of something that is... well maybe not interactive but still breaks the "just an HTML document" page metaphor that plays nice with such things.
Now, personally I find the UI compromises it takes to make endless scrolling happen are abhorrent (like not being able to ctrl-F or export the page content). But other people obviously like them.
Is that the kind of thing you'd be giving up as you move to a platform like HTMX?
Maintaining a code path with a snippet for each possible viewing experience, back when REST was all the rage, was a bit obnoxious though.
I get lazy-loading for pages that contain a ton of media of which the user will likely only see a fraction. But some people seem to think infinite scroll is a good idea with text based content as well..
That's how interactivity has always worked.
Server-side render everything + ship interactivity via js on top of it.
I feel like most of the web developers forgot that React/Vue/etc solve a specific problem: single-page application.
This is a very narrow and specific problem: navigating from page to page not causing full page reloads.
But the web has changed since SPAs were needed in major ways:
1. the average device and connection is insanely faster than 10-15 years ago. Receiving and rendering content is not the problem it once was in the age of 2g connections and limited hardware mobile devices. Even the very low end phone from few years ago has nowadays 0 problems loading and parsing few hundred kbs of html and js.
2. Web technologies moved at all levels. Server technologies, cloud ones, browser ones. React and company are solving problems that barely belong to the modern web.
In short: today you can have an app-like, spa-like experience even out of fully server-side rendered applications. While also benefitting from shipping much less code to the client.
There's a very minimal amount of websites and applications out there that benefit, and aren't crippled by these rendering libraries: those that vastly leverage offline capabilities of the platform and want to keep working regardless of internet being available. I'm talking the notions and linears.
But bar from those, still working with this React slop is just not good for the user. Even multi billion companies hiring leetcode ninjas can't get acceptable user experience and performance out of those libraries, and it's not a coincidence: they keep forcing the wrong tool for the wrong problem.
They keep living in 2018 and at the end of the day the only excuse for those react/angular+tailwind slop is that there's an entire generation of developers that doesn't know anything else and has long lost any proper engineering skill (if there is any) in finding the right tool for the problem. This is a familiarity issue, not sound technical decisions. It's the "you can't go wrong with oracle/mysql/ibm", but actually you can, and it shows.
What's worse: it's incredibly cheap to experiment different solutions and approaches via LLMs in 2026, but people keep slopping the same monstrosities.
Using the same reasoning, I’d come to the opposite conclusion. Devices are much faster than they were 10-15 years ago, but network latency is still roughly the same because physics limits how much it can improve. So reducing network round trips matters more than ever, which favors SPAs over MPAs.
That's not true. They solve the problem of requiring separate code paths for interactive elements - one for initial render and another for updates.
Previously the server rendered the initial HTML and then client JS updated it after e.g. user clicks a button, but with React/Vue/etc., it's merged into one code path because HTML is derived from the current state.
> Even multi billion companies hiring leetcode ninjas can't get acceptable user experience
Modern UX is bad because many front-end programmers fundamentally just do not care about the UX, it's not about the specific tech they use.
Because you can always just put a mini residential building inside a parking garage if you need tenants.
Need an office? Just put a mini office in the parking garage. It's perfect.
Parking garage construction is so clean and simple. Other buildings are too complex! From now on, I'll only build parking garages.
(Then I guess you have the opposite example of the residence without the parking garage! It's a great place to live, but there's nowhere to store my things and park my car!)
However for some really complex mini apps, that's another story. But for the rest of those CRUD pages, you can go simple server side rendered.
It works quite well, and saves me an awful lot of tedious mucking about with other stuff.
I have been writing frontends since early 2000. So I have seen it all, from activex being shinny to jquery, mootools, backbonejs, angular 1.0, php, Java Spring, Go. I looked into htmx and it is very much a second attempt at angular 1.0, which I did use for some good half decade as that was the best option at the time, but sooner or later, you get sick of stuffing "little codelets" inside attributes all over the place, which is exactly what htmx does.
If you want to understand what htmx is going to look like at scale, look at angular 1.0 projects.
If it's a traditional React SPA, you can use a versioned bundle with immutable caching and each user only has to download your bundle when you release an update or they use a new browser. You will need to think about how you handle spikes in traffic when you do a release, but even then users won't all load your page immediately after you bump the version.
Most of the time (existing user using the same browser, no update) there will be precisely zero traffic to your servers to load the React app except for the initial HTML skeleton.
For an app using SSR it's more complicated and I'll admit I don't know how that works at very high load
Besides the memes, it is absolutely not hype-driven, but hypermedia driven. It asks the question: could HTML be even more powerful than it already is?
The creators of HTMX even want to standardize core ideas of HTMX into the official HTML specification: https://triptychproject.org/ Please read this and reply when you still think it's hype.
Hypermedia is what to web apps what XML is to programming languages. We have tried HTMX as a concept many times over, there is nothing new here, and like everything declarative, sooner or later it will fall short and you're going to reach for escape hatches and what not.
And the features specified in that project is nice to have, in the same way that it is nice that we have Date Pickers or other advanced input features, but it is never going to replace React-like frameworks.
Again, the reason we have finally stabilised on JSX is because you can't really "Declare" away HTML or sophisticated data and event management, Google really really tried that with Angular 1.0, and we know it doesn't scale.
steer clear.
The creator of HTMX would disagree with you (as the person who described and coined the term REST)
https://htmx.org/essays/how-did-rest-come-to-mean-the-opposi...
It was only the right side results of the experience that needed to be refetched when the form changed.
But in HTMX, you are expected to just send the whole dang thing back for every change. Makes sense for a like-button, a dialog, or even a classic form submission.
My refactor cut the HTML response by about 70kb and that made a big difference. The form became fully static and I had a simple Alpine data component that would just sync that state with the URL. Still used a GET on the form and still used the same partial endpoint. Just way less HTML to wait for.
In fact, the complexity of Chat Messages and Notifications synchronisation is one of the prime reasons why React exists.
Which makes it even more ironic that in the 10 years since React was announced in 2013 and when I quit Facebook, they never managed to fix the original state synchronization bug that they demonstrated in the original React announcement for more than a few months at a time at best.
I'm sure technically it wasn't a singular bug over the years and the causes were complex, but React did fuck all in relation to the unread message counter being correct.
SSE is perfect for chat/notifications and is a lot simpler than syncing. The server pushes updates to the client doesn't require a sync. SSE is also more performant since it can keep the connection open. The example below is on a $10 vps last I heard.
If you want basic interactivity, HTMX is fine.
If you want something more (like sortable) I'd reach for lit instead of adding the Sortable.js dependency, something the HTMX docs suggest ( https://htmx.org/examples/sortable/ )
And yes there are also some scenarios that you want to control the whole routing and rendering stuff, then go for Solid, React, Angular, Vue, whatever.
There is no golden rule to all of this but finding out what makes sense is a fun exercise.
But interactivity in largely content-driven sites doesn't necessarily preclude HTMX. You can do a bunch of fancy interactive stuff with HTMX and some minimal JS. Drag and drop sort is a tiny drop-in JS snippet. Skeleton loaders and spinners are supported out of the box in HTMX as is infinite scroll as others pointed out.
I'm not saying Next.js, Solid Start, Remix, Nuxt or some other hybrid server and client rendered model doesn't work for forum software. Or that even client only rendering can be a valid approach if you accept the trade-offs (no curl-ability). People should use whatever they feel productive building with.
But if your backend is already a non-JavaScript server rendered framework like Django then HTMX makes a ton of sense to add some interactivity.
Otherwise, I'd recommend HMTX even if you have a super flexible JS backend.
That said, this is all baseless thinking without seeing any requirements and as I said before, trying to choose the right abstraction is the hard but fun part.
"Fun" as in you're years into a project before you fully accept that you made the bad call.
People weren't cancelling the page load and surfing back anymore. You cannot argue that React is extremely heavy and also adds a lot of time between first render. Test a normal HTMX page in Lighthouse and then convert it to React and you'll see how drastic things are.
People use React today out of habit rather than because it's the best tool for the job.
But while first render is of course important, now everything else is more expensive with htmx because you have to generate HTML on the server, on top of your usual fetch pipeline. Yes, for small traffic it makes little difference, but once you have >10k/s requests, things start to cost.
HTMX tools are simpler while providing straightforward thinking around html fragment caching, whole page caching. React just gives a runaround way of doing everything. I'll remind you React was built so that people could continue watching a streamed video while they navigated on a page. Most pages aren't streaming anything.
What does React have to do with loading times of a page? React does not contribute to that at all other than having to download the JS runtime, which with Preact is 3KB.
> You cannot argue that React is extremely heavy and also adds a lot of time between first render
Yes you can! In no world would React itself add enough render time to make people navigate back, even if running on a computer from the 90s!
Again, the issue with htmx is that it pretends like it is not a framework, when in fact, it is just a second attempt at angular 1.0, with even more naive assumptions about web apps.
I have no idea what this means. The World Wide Web itself is quite literally hypermedia.
The fact that a lot of front-end frameworks appear hell bent on ignoring this fact doesn't make it any less true.
> Again, the reason we have finally stabilised on JSX is because you can't really "Declare" away HTML or sophisticated data and event management...
You may have stabilised on JSX, "we" have not. React is one way of building web applications. It's appropriate for a certain subset of highly interactive SPAs, and completely inappropriate for many other things.
Just about any reputable sources puts the combined market share of React, Vue, Angular 2+, and Solid well above 80%.
So I am not sure what "we" you are talking about.
I have no idea if that's accurate, but let's assume for a moment that it is.
- Vue supports JSX, but it is not the default.
- Angular does not support JSX.
- Svelte, which you neglected to mention, does not support JSX.
- Solid does indeed use JSX, as of course does React.
So two out of the five main SPA frameworks don't even support JSX, and another doesn't typically use it.
As I said, you may have stabilised on JSX, "we" have not.
JSX just feels weird.
I think he's saying that the Web is hypermedia plus Javascript, rather than just hypermedia. Which means, from the standpoint of the power of hierarchy, that it's Javascript, which also happens to use some hypermedia. Even if you use only a tiny bit of a Turing-complete language, Turing-completeness defines what your system actually can do.
This makes the Web fundamentally awkward, because most of the behavior is defined by the hypermedia part, and users have inconsistent expectations for what happens when you break the illusion.
HTMX works by extending the illusion. JSX works by breaking the illusion earlier rather than later.
SPA’s tend to require more network round trips, sometimes significantly more based on the architecture.
The UX for an SPA can indeed be excellent, but the average SPA is worse than the average MPA, in my opinion. It’s just much easier to break things in many subtle ways in an SPA.
But for more details, see the sibling comment:
If it's a sub-function of a bigger ensemble that mostly dies something else then it's fine to call them “mini apps”.
For instance a bank app may contain a text editor for the internal messaging feature of the app, that would count as a min-app in that context.
This is untrue and is weird misunderstanding on your part. Most of modern UX and UI is designed to fit the needs of the customer for the frontend team/dev: the product or marketing or sales department. The user of the website is very rarely the customer for the dev.
That's the definition, I didn't made it up.
And those libraries solve that.
Your other definition, can be fully implemented server side, Elixir's Phoenix does that. But it's not the only one.
Your prior claim was that they only solve one problem, not about a definition. They solve several problems:
- less data can move over the wire compared to sending a full page every time (this is in your definition, and is definitely not always true, as sometimes REST APIs can return far more data than what's needed to render the screen)
- the same REST/data API can serve your website and other consumers, e.g. any native mobile apps or third party API consumers
- you can write tests for your frontend, e.g. you can unit test components' behaviour and style in isolation
- the frontend can also work, or at least respond usefully to the user, when the network or the backend are not working
In a reasonably complex application, each interface ends up needing its own shape of optimized APIs. Otherwise the clients become very complex munging/reshaping data or making too many API calls, throwing away a lot of data etc.
According to what you said, CSS also solves the problem of making an SPA because you could "switch pages" by changing CSS class names on `body`. But if someone asked what CSS is you're not gonna answer "you use CSS to make an SPA", even though _technically_ you can use it to make an SPA.
The only thing the 2 have in common is the use of HTML attributes for functionality. Completely different on every other axis that matters.
In another comment, you mentioned State Management. If this is on your mind then you are using htmx wrong. You should not be managing any client side state with htmx. State is on the server or in your database. Interactions on the client should immediately reflect the updated server state. If you have separate state on the client that needs to be managed, you are going to have a bad time regardless of framework.
Say you have a huge Google Ads budget burning up on those landing pages. Make one version with React and one with HTMX. Open Lighthouse and see performance difference. First page renders are much slower with React. Page load is slower and this makes a huge difference in mainland USA where people are still on 3G speeds on their budget phones.
"High traffic" might be the simplified view here and it took me a few page memos to explain it to my lead but the gist is that React is slower, takes longer to load and most of what it is used for is easily done with plain HTML and sprinkles of JS as needed. Unless you're streaming sound or video like you do on Facebook pages it's really not the right tool for the job in a majority of cases.
The gist of it continuously seems to be that people who think HTMX Great, know very little about good frontend engineering.
But I will answer your main questions since your comment seems to be in good faith.
Just like angular 1 choked on complex, high-frequency UIs, HTMX is going to follow the same faith. Github uses a similar approach to HTMX and you will find that even for such a simple system, their notification indicator elements on the same page are often out of sync.
The reasons for this is that State Management is hard, which is why the reconciliation loop of React-like frameworks and shadow-dom does away with pushing complex state management to user application and handles it systematically.
Of course, this doesn't mean htmx is useless, for a specific case of web apps, it is good enough, but the problem is that often times, you start with "specific cases" and your application grows overtime.
So when you consider that and the cost of doing htmx vs React or any react-like framework, picking htmx is only reasonable if it is the only option you're comfortable with and rarely on merit.
Yeah, because clients never need to keep track of state like which block is expanded or which tab is selected or that the previous page was or what the user is currently typing...