The part that always gets me is when people choose Js for the backend. Like I get that it’s the default thing that works on the frontend, so there’s some rationale why you might not want to transpile to it from another language. On the backend though, there are so many far better option, why would you willingly go with Js, especially given that you’re now forced to do all your IO async.
I moved from primarily ASP.Net Core backends, which is a hell of a great backend framework btw, to NestJS. Not my choice. I do what the people who sign my paychecks ask for.
I cannot begin to fathom why anyone would willingly choose JavaScript for backend. TypeScript helps a lot but there are still so many drawbacks and poor design decisions that make the developer experience incredibly frustrating. Features that are standard in ASP.Net Core, Django, or other common backend frameworks just don’t exist.
Also, don’t get me started on GraphQL. Sure, it has performance advantages for websites of a certain size and scale. But 99% of the websites out there don’t have the challenges that Facebook has. The added complexity and development cost over REST is just not worth it.
Yeah, gql in particular is a problem looking for a solution in most cases. It makes sense for facebook because they have people building frontend apps for their marketplace, and those apps need all kinds of weird combinations of data. However, this isn’t the situation for most apps where you have a fairly well defined set of calls you’re doing.
Server side rendering looks like it could be useful. I imagine SSR could be used for graceful degradation, so what would normally be a single page application could work without Javascript. Though, I’ve never tried SSR, and nobody seems to care about graceful degradation anymore.
Most pages tend be just documents and fairly simple forms. Making SPAs and then having to worry about SSR is just making a Rube Goldberg machine in most cases. I think something like HTMX is a much better approach in most cases. You keep all your business logic server side, send regular HTML to the client, and you just have a little bit of Js on the frontend that knows how to patch in chunks of HTML in the DOM as needed. Unless you have a highly interactive frontend, this is a much better approach than making a frontend with something like React and adding all the complexity that goes with it.
No one cares about graceful degradation anymore. But you can sell management on SEO. Page performance is a key aspect of search engine rankings and server-side rendered pages will almost always have a much faster initial load than client-side rendered.
No I meant having to do async as opposed to having threads like you would in Java for example. In vast majority of cases a thread pool will work just fine, and it makes your code far simpler. Typically, Java web servers will have a single thread that receives the request and then dispatches it to the pool of workers. The JVM is then responsible for doing the scheduling between the threads and ensuring each one gets to do work. You can do async too, but I’ve found threads scale to huge loads in practice.
The part that always gets me is when people choose Js for the backend. Like I get that it’s the default thing that works on the frontend, so there’s some rationale why you might not want to transpile to it from another language. On the backend though, there are so many far better option, why would you willingly go with Js, especially given that you’re now forced to do all your IO async.
I moved from primarily ASP.Net Core backends, which is a hell of a great backend framework btw, to NestJS. Not my choice. I do what the people who sign my paychecks ask for.
I cannot begin to fathom why anyone would willingly choose JavaScript for backend. TypeScript helps a lot but there are still so many drawbacks and poor design decisions that make the developer experience incredibly frustrating. Features that are standard in ASP.Net Core, Django, or other common backend frameworks just don’t exist.
Also, don’t get me started on GraphQL. Sure, it has performance advantages for websites of a certain size and scale. But 99% of the websites out there don’t have the challenges that Facebook has. The added complexity and development cost over REST is just not worth it.
Yeah, gql in particular is a problem looking for a solution in most cases. It makes sense for facebook because they have people building frontend apps for their marketplace, and those apps need all kinds of weird combinations of data. However, this isn’t the situation for most apps where you have a fairly well defined set of calls you’re doing.
Server side rendering looks like it could be useful. I imagine SSR could be used for graceful degradation, so what would normally be a single page application could work without Javascript. Though, I’ve never tried SSR, and nobody seems to care about graceful degradation anymore.
Most pages tend be just documents and fairly simple forms. Making SPAs and then having to worry about SSR is just making a Rube Goldberg machine in most cases. I think something like HTMX is a much better approach in most cases. You keep all your business logic server side, send regular HTML to the client, and you just have a little bit of Js on the frontend that knows how to patch in chunks of HTML in the DOM as needed. Unless you have a highly interactive frontend, this is a much better approach than making a frontend with something like React and adding all the complexity that goes with it.
No one cares about graceful degradation anymore. But you can sell management on SEO. Page performance is a key aspect of search engine rankings and server-side rendered pages will almost always have a much faster initial load than client-side rendered.
You really should be doing your IO async. Do you specifically mean callback hell?
No I meant having to do async as opposed to having threads like you would in Java for example. In vast majority of cases a thread pool will work just fine, and it makes your code far simpler. Typically, Java web servers will have a single thread that receives the request and then dispatches it to the pool of workers. The JVM is then responsible for doing the scheduling between the threads and ensuring each one gets to do work. You can do async too, but I’ve found threads scale to huge loads in practice.
Green threads are functionally the same, especially in languages that can preempt.
Sure, but the scheduler figures out the scheduling automatically so you don’t have to worry about stuff like blocking.