Monthly Archives: August 2011

JSIO Performance Results

Posted on by
Reply

Below are two graphs plotting the number of images on a page against the total KB transferred by Chrome 13 for both a page with JSIO and without JSIO (which I, for some reason have called a “plain” page).

I build websites for a living, so I have a lot of folders of images from various websites lying around which form the two data sets for the graphs. I’d like to think of them as a typical set of website images, but you may argue that they are typical for the websites that I build.

I knocked up a script that when given a folder of images, will output a “plain” HTML page, a JSIO HTML page and a JSIO resources file with the following numbers of images on them: 2, 4, 6, 8, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000 and 1100. Before generating any of the pages, the order of the images were randomised to better represent a web page’s inclusion of different images for different purposes.

Chrome 13 was used with the developer toolbar to determine how many requests were being made and the actual KB transferred (with slight rounding errors) by the browser. In my opinion I would say that Apache was set up to serve image and javascript content typical of many servers on the internet in that images were not gzipped but javascript files were; the register, bbc, for example.

What is interesting to note is that the number of bytes transferred by JSIO are consistently lower than the bytes transferred by a plain HTML page…but not by a lot. The file size inflation due to base64 encoding seems to cancel out a lot of bytes the gained from making fewer requests. However, it is obvious that JSIO won’t actually lose you any bytes, but it isn’t going to save you a substantial amount. That said, saving bytes isn’t the only benefit of JSIO, remember that one of JSIO’s goals is also to alleviate some of the problems surrounding the creation and use of image sprites – which it does also do very well.

The graph below is just for fun, it plots the number of requests made by JSIO pages against the number of requests made by plain HTML pages. Note the JSIO pages make a constant 4 requests (index.html, jsio.js, jsio-resources.js and jsio.gif) independent of how many images are on the page:

I want my Snow Leopard functionality back!

Posted on by
8 Replies

Lion’s Mission Control is the closest thing I have to Snow Leopard’s Spaces. Generally, I’m fine with Lion. I’m used to the reverse scrolling and can just about deal with not knowing what I can and can’t scroll (although that is starting to grate), but the following things really affect me as they are “killer features” that I now don’t have:

  • No see “all windows in all spaces”. In Snow Leopard I had bottom left as my spaces hot corner and bottom right as my all windows hot corner. What was awesome was that I could trigger spaces at the bottom left, and then trigger all windows at bottom right – which meant I could see all windows in all spaces! Mission Control shows me all windows in the current desktop, but not all windows in other desktops…grrr
  • No moving windows between spaces/desktops. In Snow Leopard I used to be able to move windows between spaces, I cannot, in Lion, move windows between desktops…double grrr
  • No rearranging desktops in the order I want them. Also note that “Automatically rearrange spaces based on most recent use” is completely useless. I really don’t want Lion to move my furniture around…how am I ever going to find stuff if it isn’t left in the place I put it?

JSIO URLs

Posted on by
2 Replies

The clever bit about JSIO is the placeholder URLs – i.e. the “jsio.gif#…” that goes in your image src attributes and background-image CSS properties. Everything after the “#” is the filename of the image that should be displayed. It doesn’t have to be the filename at all, it could just be a single letter or number or symbol or whatever. It doesn’t matter to JSIO. To JSIO, everything after the “#” is just a key into the resources object that holds data uri encoded images. As long as the key is unique (which it will be if you use filenames, since no two files in the same directory can have the same name), JSIO is happy.

Using filenames as keys into our resources object is beneficial to humans. Much more beneficial than coordinates in a sprite, simply because (hopefully) the filenames are meaningful; they describe the image. In comparison to sprites, the JSIO resources object also makes maintaining your image data much easier as it is trivial to add or remove images without having to move other images around within the sprite (and consequently all coordinates referencing your moved image).

Another benefit of using filenames as keys into the JSIO resources object is for fallback. If JSIO detects your browser is IE7 or lower, it’ll strip out “jsio.gif#” leaving just your image key as the image src, which is hopefully a valid URL to the original image. Also, if JSIO detects you’re running IE8 and the image data is larger than 32KB it’ll do the same thing*.

* …but not yet in v1.0.0 alpha

Since the image key is after the “#” (it is the URL “fragment”), your browser won’t send multiple requests for the 1*1px jsio.gif file – it’ll just send one request, cache the response, and use it again. By the way, the jsio.gif image is just a transparent 1*1px gif (for maximum efficiency), but it could be an “spinner” image or something, which is shown temporarily whilst the JSIO resources file is downloaded.

—

This post is about JSIO – JSIO is a tiny library that allows you to make fewer requests to your server by packaging all your site image data in a JavaScript file in data uri format. The official site for JSIO can be found here: jsio.freestyle-developments.co.uk. You can read more about why I started this project here.

JavaScript Image Optimiser (JSIO)

Posted on by
2 Replies

So, I’ve embarked on a new mini project. It is kind of inspired by image sprites.

Image sprites are a great idea, but come with a whole bunch of issues that make them a bit of a pain to work with.

Firstly, most of the time you have to use markup to create an element in html to “hold” the image you wanted to display from your sprite. This is because if you actually set a sprite as the background image for a large html element you’re likely to see other images in the sprite as well. Because of this, you actually lose useful functionality that CSS gives you, like the ability to position, repeat and scale an image. Also, the markup you’ve created to hold the image exists for style purposes, which is bad.

Secondly, sprites can be a massive ball ache to maintain. If you’ve closely packed your images in a sprite for maximum efficiency and then one of your images needs to change size, you’re either going to have to move ALL images surrounding the image you have to update (and obviously then change all background-position properties for the images you’ve moved) or leave a space and put the updated image in a new position in your sprite.

How do you know which images in a sprite are used and which ones are dead? Since your images are referenced by coordinates, this sort of clean up becomes a nightmare and is actually a bit lot of a nightmare to create image sprites in the first place.

The goals of the project are to

  1. Reduce the number of http requests (and their associated header traffic) to the server and hence reduce the time it takes to load all images on a website and bandwidth footprint the site requires
  2. Create a solution that’ll alleviate some of the problems surrounding the creation and use of image sprites
  3. Do something cool

The JSIO project website has a pretty good explanation of how it works so I won’t bore you with the details here. However, as a brief overview, it packages all your image data in data uri format and you reference particular images by their filename rather than their coordinates.

The site actually uses the HTML5 file api to generate your resources file for you, which makes creating and maintaining your JSIO “sprite” really really easy.

I read *somewhere* that data uri encoded images can be up to 1/3 larger than corresponding image files, however with gzip encoding they can be only 0-3% larger (or less). My thesis is that for a site with many small images, JSIO could be more efficient and easier to maintain than having separate files or even an image sprite.

…I’m yet to prove or disprove this and I’ll be conducting some tests whose results I’ll post up here (even if they do prove JSIO to be useless).

Disclaimer: This is the first ever ALPHA release of JSIO – it works on the latest Firefox and Chrome but I haven’t even checked it in IE yet. It probably won’t work in IE yet. Also, the website needs some work for optimal display on mobile devices.