embedded vs linked JS / CSS

The one thing the other answers didn't touch on is developer efficiency. If it's easier to put it inline, and there's no immediate performance requirement/concern, then do that. There is real business value to "easy", and it trumps eventual or non-existent performance concerns. Don't prematurely optimize.

The advantage of linked JS files is that they can be cached by the browser and loaded from local disk or memory on subsequent pages.

The advantage of inline JS is that you might have fewer network requests per page.

The best compromise is usually a small number of linked JS files (one or two) that consist of a mininified combination of all your JS so they are combined into as few files as possible and as small as possible.

Getting the benefits of local caching far exceed the extra parsing of a little JS that might not be used on some pages.

Embedded JS that makes sense (even most of your JS is in linked files) is the settings of a few JS variables that contain state that is specific to your page. That would typically be embedded into the section of the page as it's generated dynamically at your server, different for every page and usually not cacheable. But, this data should typically be small and page-specific.

Linking a script incurs a small penalty in the form of an extra request to the server. If you keep it inline this request is not made and depending on the situation you may get a faster loading page. It makes sense to inline your code if:

• it is very small
• it is dynamcally generated since then you won't get the benefits of caching anyway

In the case of google and facebook you're most likely seeing inline javascript because it's being generated by server side code.

Other answers have already mentioned the advantages of caching with external JS files. I would almost always go that way for any library or framework type functionality that is likely to be used by at least two pages. Avoid duplication when you can.

I thought I would comment on "inline" vs "embedded".

To me, "inline" means mixing the JS in with the HTML, perhaps with several separate <script> blocks that may or may not refer to each other, almost certainly with a number of event handlers set directly with HTML element properties like <div onclick="..."/>. I would discourage this in most circumstances, but I wouldn't get too hung up about it for occasional uses. Sometimes it's simply less hassle and pretending otherwise just wastes time that you could spend on more important issues.

I would define "embedded" as having (preferably) a single <script> block in the head or at the end of the body, with event handlers assigned within that block using document ready or onload function(s). I see nothing wrong with this for functions specific to one page, in fact I tend to prefer this over an external file for that purpose if it's only a small amount of script and I don't care about caching it client-side. Also if the page is generated dynamically and something in the JavaScript needs to be generated on the server it is generally much easier to do it if the script is on the same page.

One last note on external files: during development watch out for IE's tendency to "over cache". Sometimes while testing I've made some small changes to an external file or two and pulled my hair out wondering why it didn't work only to eventually realise that IE was still using an old cached version. (On the one hand of course this is my fault, but on the other I know a lot of people who have fallen victim to this from time to time.)

All the answers above are very good answers. But I would like to add my own based on 15 years of web dev experience.

ALWAYS use linked resources for both CSS and ECMAScripted resources rather than inline code as linked content is cached by the browsers in most cases and used across potentially thousands of pages over hours and days of interaction by a user with a given domain online. The advantages are as follows:

1. The bandwidth savings using linked scripts are HUGE as you simply deliver less script over the wire over the user experience which might use the cache for weeks.
2. There's also better cascade of CSS, as embedded and inline styles override, by weight, linked styles causing confusion in designers.
3. There is avoidance of duplicate scripts which happens a lot with inline scripts.
4. You reuse the same libraries over many pages with cache control on the client now possible using versioning and date-based querystrings.
5. Linked resources tell the browser to preload all resources BEFORE initializing scripts of styles in the DOM. Ive seen issues related to this where users pressed buttons prior to pre-processing of date-times in time sheet apps by scripts causing major problems.
6. You have more control over script and CSS libraries by all developers in a project, avoiding polluting your app with hundreds of poorly vetted custom scripting in pages
7. Its very easy to update libraries for your users as well as version linked libraries.
8. External script libraries from Google or others are now accessible. You can even reuse your own linked libraries and CSS using linked resources.
9. Best of all there are processing speed gains using cached client-side resources. Cached resources out perform on-demand resources any time!
10. Linked scripts also enforces style and layout consistencies instead of custom layout shifts per page. If you use HTML layouts consistently, you can simulate flash-free page views because cached CSS is used by the DOM across web pages to render pages faster.

Once you pull in linked resources on the first domain request/response the user's experience is fast and server-side page delivery means the DOM and HTML layouts will not shift or refresh despite numerous page views or links to pages across the site. We often then added limited custom page-level embedded style and script resources as needed to the cached linked stack of libraries on a page level if needed for a narrow range of customizations. Global variables and custom CSS can then override linked values. This allows you to maintain websites much easier without guesswork page-by-page as to what libraries are missing or already used. Ive added custom linked JQuery or other libraries in sub-sections to gain more speed this way, which means you can use linked scripts to manage custom subgroups of website sections, as well.

GOOGLE ANGULAR

What you are seeing in Google's web presence is often implementation of Angular's very complex ES5 and ES6 modular scripted cache systems that utilize fully Javascripted DOM manipulation of single page applications using the scripts and embedded CSS in page views now exclusively managed in Angular(2+). They utilize elaborate modules to load on-demand and lazy load components and modules with HTML/CSS templates pre-loaded into memory and from the server behind the scenes to speed delivery of news and other pages they manage.

The FLAW in all that is they demand browsers stream HUGE megabytes of ECMAScript preloaded with HTML and CSS embedded into these webpages behind the scenes as the user interacts with these cached pages and modules. The problem is they have HUGE stacks of the same CSS and scripts that get injected into multiple modules then parts of the DOM which is sloppy and wasteful. They argue there is no need now for server-side delivery or caching when they can easily manage all that via inline style and script content downloaded via XMLHTTPRequest hidden WebAPI calls to and from the server. Why download all that and rebuild and store inline pages in memory constantly when a much smaller file linked from the page would suffice?

Honestly, this is the sloppiest approach to cache management of styles, content, and CSS I have seen yet in web dev frameworks, as it still demands huge megabytes of scripts just to parse a simple news page with a few lines of text. Someone at Google didn't really think that framework through lol. Its wasteful of bandwidth and processing in the browser, if you ask me, and overkill. Its typical of over-engineering at these bloated vendors.

That is why I always argue for linked CSS and scripts. Less code and more content is why these frameworks were invented. Linked and cached code means SIMPLER, OLDER models have worked better using the fast delivery of smaller markup pages that cache tiny kilobytes of linked ECMAScript and CSS libraries. It means less code is used to display content. The browser's relationship with the server now is so fast and efficient today compared to years ago that the initial caching of these smaller linked files directly from the server (rather than giant inline pages of duplicate scripts yanked down via Web API in Angular on every page view) means linked resources are delivered much faster over the initial visit of a typical web domain visit.

Its only recently the 'script kiddies' have forgotten all this and so have started going backwards to a failed way of using local embedded and inline styles and scripts which we stopped using 20 years ago for a reason. It is a very poor choice and shows inexperience with the web and its markup and content model by many new developers today.

• Stokely