How can you / is it possible to avoid making server request per web component?
Background
So, with javascript a lot of people are now using the so called AMD model of module loading, in which all the modules are separate javascript files and get included as required. eg.
- main.js
-- module1/mod1.js
-- module2/mod2.js
-- module3/mod3.js
This results in multiple requests to the server, one for each javascript file. This is considered harmful to app performance, particularly in mobile as the number of javascript modules increases.
As a result tools such as require.js provide a compiler that will automatically trace dependencies and generate a single javascript file with all the modules in it; effectively reducing the request overhead to a single file (typically app-min.js).
Web components
Every web component sits in an external file island and gets imported using a link tag in the header:
<link rel="import" href="elements/image-gallery.html">
<link rel="import" href="elements/social-media.html">
<link rel="import" href="elements/pinmap.html">
<link rel="import" href="elements/nav-menu.html">
If you're using polymer, you can use the vulcanize tool to combine polymer components (http://www.polymer-project.org/articles/concatenating-web-components.html), but that's because (as I understand it) polymer is a javascript framework that loads web components dynamically from "polymer-element" tags.
Is there an equivalent way of doing this with 'vanilla' web components that do not rely on a frame work such as x-tag or polymer?
Vulcanize
The Vulcanize tool can be used with vanilla web components (i.e. HTML imports) as well as with the Polymer framework. It has some drawbacks however (build setup and nasty bugs such as incorrect parsing of SVG etc.).
HTTP 2.0 (SPDY)
IMHO, the best option to get rid of the multiple request roundtrip problem is to use SPDY (HTTP 2.0). This allows a single physical roundtrip for all requests responses without requiring you mess up your nice and clean code. It works for HTML, javascript, images and web components alike. I.e. you can keep your HTML and Javascript nice and modular without the penalty of request hell. This is the most generic approach and provides the cleanest and most effective solution. The downside is that it requires a modern server side application hosted in a modern web server. So if your server supports it, this is the silver bullet of avoiding request hell without arcane setups and time consuming patterns.
Manually
If none of these approaches fits your bill, you can violate all things beautiful and manageable and optimise things by manually concatenate content in the same way you would with vanilla web content. There is little magic going on when it comes to web components. The biggest thing is actually that HTML document fragments are often kept in separate files (using HTML imports), but this is not actually a requirement but rather a best practise. HTML imports can be used without web components and web components can be used without HTML imports. Web components typically use document fragments and Javascript and it is convenient to wrap both in an import. To avoid the extra requests, you can simple define multiple fragments in a single big files. You do this by using the template element for each fragment in a single HTML file just could easily just merge multiple scripts together in a single file. This is basically what the Vulcanize tool does as it does not actually try to interpret the code being concatenated. No need to say that this approach replaces request hell with maintenance hell.
Related
Current situation:
We have 1 application that generates sass files, which is then obviously compiled into CSS and uploaded to 1 CDN location. This is done by "us" (the design team).
There are multiple applications that use this css CDN. This is done by the devs, who just import said CDN, ready to use. (To clarify: the solution can not involve anything on the dev-end.)
This one CDN contains
basic stuff that all applications need (color variables, mixins, etc)
application specific styling: meaning there are rootclasses (.application-1 {}, .application-2 {},...) that contains everything for that specific application, so it doesn't interfere with anything else
Now as you might imagine, that one CDN is becoming very bulky and every application is loading all styles for all applications which is unnecessary heavy.
So what we want:
1 CDN with the general stuff that every application will use
multiple CDNs for application-specific stuff
The problem:
Since it's SASS that we use, I can't just say to gulp to generate separate files and be done with it because the specific SASS relies on the mixins and variables of the general SASS.
I hope this is clear, it's a tad difficult to explain properly. I'm hoping someone has had this same issue and can tell me how they solved it.
Recently we have experienced the problem of browsers caching the css files, so that when we move code to another web server we get unexpected and unwanted results.
Having read various solutions it occurred to me that the css file could be included in the template. Rather than write
<link href="/static/css/parentx.css" rel="stylesheet" type="text/css">
we could put
<style>{% include "static/css/parent.css" %}</style>
which would embed the css in the html. It has the added benefit of reducing the number of calls back to the server for the css file but the disadvantage of not caching the css at all and not allowing the web server to provide the static files. However, these files are quite small, typically about 12K.
I propose this as a temporary measure as I have just discovered the existence of asset management plugins that look to do a better job.
I have never seen the use of the 'include' template command to get css (and javascript) files so perhaps there is a drawback of which I am unaware?
Not caching the CSS file at all will slow page loading though it's unnecessary.
Cache your CSS as long as possible and necessary and when you need to invalidate cache, a sure way is to load another file, I mean the same file with another name. It could be parent1391234565.css or parent-v003.css or parent130307.css or parent-4e6ca1437f2d.css. If you told browsers to download parent-v002.css, you're pretty certain they don't have parent-v003.css in their cache!
When I creare stylesheets for my ASP.NET MVC 4 web site everything works great when in debug/development mode.
As soon as I deploy the web site on IIS, in release config, some parts of the css are not being applied to the elements since they are not present at all in a single minified .css file that is being added to the page.
Making my declaration more specific - e.g. including id > class or stuff like that ususally solves the problem, but what are the general rules for writing css styles so that they are served to the client and are not filtered out by ASP.NET minification?
If you're talking about ASP.NET bundling, it will bundle the CSS files in alphabetic order by default. One simple way to make sure the files are always rendered in the correct order is to use a prefix on the filename, e.g.:
01.first-file.css
02.second-file.css
03.third-file.css
04.fourth-file.css
Having said that, making your declarations more specific and therefore less dependent on the ordering of files is probably a good idea.
We can write CSS as the following types:
Inline CSS
Embedded CSS
External CSS
I would like to know pros and cons of each.
It's all about where in the pipeline you need the CSS as I see it.
1. inline css
Pros: Great for quick fixes/prototyping and simple tests without having to swap back and forth between the .css document and the actual HTML file.
Pros: Many email clients do NOT allow the use of external .css referencing because of possible spam/abuse. Embedding might help.
Cons: Fills up HTML space/takes bandwidth, not resuable accross pages - not even IFRAMES.
2. embedded css
Pros: Same as above regarding prototype, but easier to cut out of the final prototype and put into an external file when templates are done.
Cons: Some email clients do not allow styles in the [head] as the head-tags are removed by most webmail clients.
3. external css
Pros: Easy to maintain and reuse across websites with more than 1 page.
Pros: Cacheable = less bandwidth = faster page rendering after second page load
Pros: External files including .css can be hosted on CDN's and thereby making less requests the the firewall/webserver hosting the HTML pages (if on different hosts).
Pros: Compilable, you could automatically remove all of the unused space from the final build, just as jQuery has a developer version and a compressed version = faster download = faster user experience + less bandwidth use = faster internet! (we like!!!)
Cons: Normally removed from HTML mails = messy HTML layout.
Cons: Makes an extra HTTP request per file = more resources used in the Firewalls/routers.
I hope you could use some of this?
I'm going to submit the opinion that external style sheets are the only way to go.
inline CSS mixes content with presentation and leads to an awful mess.
embedded CSS gets loaded with every page request, changes cannot be shared across pages, and the content cannot be cached.
I have nothing against either method per se, but if planning a new site or application, you should plan for external style sheets.
Inline
Quick, but very dirty
This is (sadly) also the only really sane option for HTML formatted email as other forms often get discarded by various email clients.
Embedded
Doesn't require an extra HTTP request, but doesn't have the benefits of:
Linked
Can be cached, reused between pages, more easily tested with validators.
You want external css. It's the easiest to maintain, external css also simplifies caching. Embedded means that each separate html file will need to have it's own css, meaning bigger file size and lots of headaches when changing the css. Inline css is even harder to maintain.
External css is the way to go.
Where to start!!??
Say you had a site with 3 pages...
You could get away with Inline CSS (i.e. CSS on the page itself, within tags).
If you had a 100 page website...
You wouldn't want to change the CSS on each page individually (or would you?!)...
So including an external CSS sheet would be the nicer way to go.
Inline CSS is generally bad. It's much easier to modify the style of a page when all the styles are located in one central location, which inline CSS doesn't offer. It's easy for quickly prototyping styles, but shouldn't be used in production, especailly since it often leads to duplicating styles.
Embedded CSS centralizes the styles for the page, but it doesn't allow you to share styles across pages without copying the text of the embedded style and pasting it in each unique page on your site.
External CSS is the way to go, it has all of the advantages of embedded CSS but it allows you to share styles accross multiple pages.
Use external CSS when:
you have a lot of css code that will make your file messy if you put it all inline
you want to maintain a standard
look-and-feel across multiple pages
External CSS makes it a lot easier to manage your CSS and is the accepted way of implementing styles.
If the styles are only needed for one file, and you don't foresee that ever changing to apply to other pages, you can put your css at the top of the file (embedded?).
You should generally only use inline CSS if:
It's a one-time formatting for a specific tag
You want to override the default css (set externally or at the top of the file) for a specific tag
To everyone in the here and now, reading after 2015, with projects like Polymer, Browserify, Webpack, Babel, and many other important participants that I'm probably missing, we have been ushered into a new era of writing HTML applications, with regards to how we modularize large applications, distribute changes and compose related presentation, markup and behavior into self-contained units. Let's not even get started with service workers.
So before anyone forms an opinion on what is and isn't feasible, I would recommend that they investigate the current web ecosystem in their time to see if some issues related to maintainability have been gracefully solved.
Pros:
Allows you to control the layout of the entire site with one file.
Changes affect all documents at the same time.
Can eliminate redundant in-line styling (Font, Bold, Color, Images)
Provide multiple views of the same content for different types of users.
Cons:
Older browsers may not be able to understand CSS.
CSS is not supported by every browser equally.
In this case, the pros far outweigh the cons. In fact, if the site is designed in a specific way, older browsers will display the content much better (on average) than if the site were managed with tables.
If you are using a server side language, why not embed CSS and use conditional programming to display it as necessary? For example, say you're using PHP w/ Wordpress and you want some homepage specific CSS; you could use the is_home() function to show your CSS in the head of the document for that page only. Personally, I have my own template system that works like so:
inc.header.php = all the header stuff, and where page specific CSS would go I put:
if(isset($CSS)) echo $CSS;
Then, in a specific page template (say about.php), I would use a heredoc variable for the page specific CSS, above the line which includes the header:
Contents of about.php:
<?php
$CSS = <<< CSS
.about-us-photo-box{
/* CSS code */
}
.about-us-something-else{
/* more CSS code */
}
CSS;
include "inc.header.php"; // this file includes if(isset($CSS)) echo $CSS; where page-specific CSS would go ...
<body>
<!-- about us html -->
include "inc.footer.php";
?>
Is there something I'm missing that makes this approach inferior?
Pros:
1) I can still cache the page using standard caching techniques (is there a caching method that takes advantage of a CSS only external file??).
2) I can use php for special case conditional formatting in specific class declarations if absolutely necessary (PHP doesn't work in a CSS file, unless I'm missing some server directive I could set).
3) All my page specific CSS is extremely well organized in the actual PHP file in which it's being used.
4) It cuts down on HTTP requests to external files.
5) It cuts down on server requests to external files.
Cons:
1) My IDE program (Komodo IDE) can't follow the Heredoc formatting to properly highlight the CSS, which makes it slightly harder to debug syntax errors in the CSS.
2) I really can't see any other con, please enlighten me!
Which is a better option: to store CSS on a separate file or on the same page?
Let's forget the fact that changing the CSS on a file makes it to apply all HTML pages directly. I am using dynamic languages to generate the whole output - so that does not matter.
A few things I can think of:
CSS on a separate file generates less bandwidth load.
CSS on a separate file needs another HTTP request.
On the other hand, if I compress the data transmission with Zlib, the CSS on the same page should not matter in terms of bandwidth, correct? So, I get one less HTTP request?
The main benefit of an external CSS file is that:
It can be used on multiple pages; and
It can be cached so it doesn't need to be loaded on every page.
So, if there is potential for reuse of the dynamically generated CSS between pages or on multiple views of the same page then an external file could add value.
There are several common patterns for dynamically generated CSS.
1. Generating a subset for a page
I've seen this occasionally. A developer decides to limit the amount of CSS per page by only sending what's necessary. I don't imagine this is the case for you but I'm mentioning it for completeness. This is a misguided effort at optimization. It's cheaper to send the whole lot and just cache it effectively.
2. User-selected theme
If the user selects a particular look for your site, that's what I'm talking about. This implies they might select a whole package of CSS and there might be a limited set to choose from. Usually this will be done by having one or more base CSS files and then oen or more theme CSS files. The best solution here is to send the right combination of external CSS files by dynamically generating the page header with the right <link> elements and then caching those files effectively.
3. User-rolled theme
This goes beyond (2) to where the user can select, say, colours, fonts and sizes to the point where you can't package those choices into a single theme bundle but you have to generate a set of CSS for that user. In this case you will probably still have some common CSS. Send that as an external CSS files (again, caching them effectively).
The dynamic content may be best on the page or you may still be able to make use of external files because there is no reason a <link> can't point to a script instead of a static file. For example:
<link rel="stylesheet" href="/css/custom.php?user=bob" type="text/css">
where the query string is generated dynamically by your header from who is logged in. That script will look up the user preferences and generate a dynamic CSS file. This can be cached effectively whereas putting it directly in the HTML file can't be (unless the whole HTML file can be cached effectively).
4. Rules-based CSS generation
I've written a reporting system before that took a lot of rules specified by either the user or a report writer and a custom report and generated a complete HTML page (based on the tables and/or charts they requested in the custom report definition) and styled them according to the rules. This truly was dynamic CSS. Thing is, there is potential for caching here too. The HTML page generates a dynamic link like this:
<link rel="stylesheet" href="/css/report.annual-sales.0001.css" type="text/css">
where 'annual-sales' is the report ID and 0001 is like a version. When the rules change you create a new version and each version for each report can be cached effectively.
Conclusion
So I can't say definitively whether external CSS files are appropriate or not to you but having seen and developed for each of the scenarios above I have a hard time believing that you can't get some form of caching out of your CSS at which point it should be external.
I've written about the issue of effective CSS in Supercharging CSS in PHP but the principles and techniques apply in any language, not just PHP.
You may also want to refer to the related question Multiple javascript/css files: best practices?
There is a method that both Google and Yahoo apply which benefit's from inline CSS. For the very first time visitors for the sake of fast loading, they embed CSS (and even JavaScript) in the HTML, and then in the background download the separate CSS and JS files for the next time.
Steve Souders (Yahoo!) writes the following:
[...] the best solution generally is
to deploy the JavaScript and CSS as
external files. The only exception
I’ve seen where inlining is preferable
is with home pages, such as Yahoo!'s
front page (http://www.yahoo.com) and
My Yahoo! (http://my.yahoo.com). Home
pages that have few (perhaps only one)
page view per session may find that
inlining JavaScript and CSS results in
faster end-user response times.
If you're generating HTML dynamically (say, from templates), embedding CSS allows you the opportunity to also generate the CSS dynamically using the same context (data, program state) as you have when you're producing the HTML, rather than having to set that same context up again on a subsequent request to generate the CSS.
For example, consider a page that uses one of several hundred images for a background, depending on some state that's expensive to compute. You could
List all of the several hundred images in rules in a seperate, static CSS file, then generate a corresponding class name in your dynamic HTML, or
Generate the HTML with a single class name, then on a subsequent request generate CSS with a rule for that name that uses the desired image, or
Do (2), but generate the CSS embedded in the HTML in a single request
(1) avoids redoing the expensive state computation, but takes a larger hit on traffic (more packets to move a much larger CSS file). (2) Does the state calculation twice, but serves up a smaller CSS file. Only (3) does the state calculation once and serves the result in a single HTTP request.
Browsers can cache the CSS files (unless it changes a lot). The bandwidth should not change, because the information is sent, no matter where you put it.
So unless the css quite static, putting it in the page costs less time to get.
I always use mix of both.
site-wide styles are in separate file (minified & gzipped),
any page-specific styles are put in <style> (I've set up my page templates to make it easy to insert bits of CSS in <head> easily at any time).
Yes and no. Use a .css file for most rules; your site should have a consistent look anyway. For rare, special case, or dynamically generated rules you can use inline 'style=""'. Anything that sticks should move into the .css, if only to make transcluding, mash-ups, etc. easier.
Keep it separate. HTML for centent, CSS for style, JavaScript for logic.