Behavior of Image ALT Attribute in IE/Fire Fox

When browsing the web, you will often notice that a tool tip displays when you mouse over certain images as shown in the image below:

Being mostly a Microsoft IE user, I assumed that the text for this tool tip was specified in the image's alt attribute, as below:

<img src="name.gif" alt="Modeling Article">

I had noticed that the tool tips didn't work in Fire Fox, but didn't pay too much attention. As I had a case where it actually mattered recently, I did some research and learned something interesting. The alt attribute is not supposed to work that way according to the official specs. IE just made it do that. Fire Fox followed the spec and did not.

The official purpose of the alt attribute is to provide alternate text in case the image doesn't load or cannot be viewed, as shown below:

The correct way to provide a tool tip, according to the specs, is with the title attribute:

<img src="name.gif" alt="Modeling Article" title="Modeling Article">

That works as expected in Fire Fox.

Of course, you want both attributes in there because search engine's make a lot of use of the alt attribute and apparently they are meant to serve two different purposes.

Always more interesting stuff to learn...

CSS-Only Rollover Buttons

I have done rollover buttons the same way for a long time now:

• Create 2 images - one "active" and one "inactive"
  
• Use a JavaScript function, launched onLoad, to preload the "active" image
  
• Use another JavaScript function, fired onMouseOver and onMouseOut, to change the source of the image object when the user mouses over it.

It always worked before. On a recent project, however, IE was giving me fits because it kept trying to reload the "active" image on every mouse over. This caused a very annoying flicker effect.

I assumed that it was some problem with my preload function and in a search for a solution I discovered a whole new way to do rollovers without a preload - without any JavaScript at all in fact. Here is the basic idea:

• Create 1 image that has both the "active" and "inactive" parts side by side
  
• Code a DIV tag that contains a link (A tag)
  
• Use CSS settings to change the offset of the background for the DIV and A tags

Since only one image is used, there is no need for a preload function. This method also works faster than the old way.

Here is an illustration of an example I put together with Home and Contact Us buttons:

To start, we need an image that contains both the "active" (yellow) and "inactive" (blue) parts for each button. Here is the home button:

The image is 300 pixels wide, which is important to know since we are going to offset this to only display half the image at any one time. It is 35 pixels high.

Now let's look at the HTML:

<div id="nav-container">
   <div id="home-button" class="nav-button">
      <a href="../index.htm"></a>
   </div>
   <div id="contact-button" class="nav-button">
      <a href="../contact-us.php"></a>
   </div>
</div>

First, we have an outside container DIV for the whole set of buttons (nav-container) that we can use to position the entire set of buttons on the page. Inside that, we have two DIV's - one for each button - that use the CSS class "nav-button". Each of those contain an empty A tag.

Now the CSS. First, we set the position of the outer container from the top left corner of the page:

#nav-container{
position:absolute;
top:25px;
left:25px;
}

Next, we will define the nav-button class:

.nav-button{
width:150px;
height:35px;
position:absolute;
top:0px;
}

Note the height and width settings: the width is half the width of the image. We set the top position from the top of the outer container, but not left as that will be different for each button.

Now we specify the settings for the anchor tags contained in the button DIV's:

#home-button a, #contact-button a{
display:block;
margin:0;
padding:0;
width:100%;
height:100%;
overflow:hidden;
background-image:none;
}

These define the link as a block that is the exact same size as the DIV that contains it. Notice that the background image is set to none. Notice also that I set this for each button's DIV. I tried doing it as a class, which would have made more sense, but that seemed to cause problems in IE8.

Now we have to specify two settings for each button: the background and left position of the DIV tag and the background of the hover state of the A tag, which is actually the same image with a different offset:

#home-button{
left:0px;
background:url('home_button.gif') top left no-repeat;
}
#home-button a:hover{
background:url('home_button.gif') -150px 0 no-repeat;
}

For the Home button DIV, we set the left position to 0, since it is the first button. We set the background image with no offset, so it will display the left 150 pixels of our image (the blue "inactive" button). Then for the Home button's anchor, we set a background for the hover state. It is the same image but it has a 150 pixel offset so that you see the right part of the image (the yellow "active" button). I don't really know why the offset is specified like that. It threw me at first, but play with it and you will figure it out.

So, when you mouse over the link you trigger the hover state, which causes the background image for the link to change from none to the offset background, which covers up the background being used for the DIV tag beneath. When you mouse off the link, the background reverts to none and the background of the DIV shows through again.

All that's left is to specify similar settings for the Contact Us button:

#contact-button{
left:160px;
background:url('contact_button.gif') top left no-repeat;
}
#contact-button a:hover{
background:url('contact_button.gif') -150px 0 no-repeat;
}

Same as the Home button except for the image name and the left position (the width of the home button plus a 10 pixel space).

Voila. No preload, no JavaScript, fewer images, better all around.

Obviously, I didn't come up with this on my own. I got the idea from Petr Stanicek's tutorial and made some adaptations. If you like, check out the demo page on my site where you can view the source and the entire CSS for this neat trick.

Pre-Packaged PHP User Login System

I have a project I am working on that requires a user sign-in/registration system. All of the user data needs to be tracked and there are requirements for limiting access to certain content based on login status, etc. It's not terribly complicated and my first inclination was to go with a WMS like DotNetNuke, which has a very robust user management system built into it. But this project will require some special modifications that would be difficult (at least for me) to accomplish with DNN.

So, my next thought was a custom PHP application with a MySQL backend. But I was concerned about the time involved in doing all the legwork just to duplicate the user account functionality that comes with DNN out of the box.

Luckily a bit of Google-ing found me a real gold mine. As is so often the case with PHP, it turns out that somebody has already done all the legwork and is willing to share. (Three cheers for Open Source!!!) Here is the URL:

http://www.evolt.org/node/60384

The download consists of about a dozen very well commented PHP pages that handle just about all the basic aspects of a web site user login system, including:

• Session-based User Login/Registration with a "Remember Me" option
  
• The ability to set different access levels for members (user, admin, etc.)
  
• An Admin page where you can view user info, upgrade/demote user levels, delete users, delete inactive users, and ban users
  
• Users can view and edit their own account information
  
• Visitor tracking that tracks both Guests and Registered Users in real time
  
• "Forgot Password" features
  
• Optional welcome email to new users
  
• Nice error handling features

He gives you the SQL you need to create the required database tables as well as some sample pages to see how the features can be integrated into your site.

It's very easy to use. I had this up and running in a test environment in about 20 minutes. The code is very well structured and commented, so I have already been able to make a number of modifications (like adding more data fields to the account profile) very quickly.

Highly recommended.

Getting an Indented List to not Indent

It seems like it should be really easy. I want a bulleted list, but with no indent. In other words, I want the bullets aligned to the left with no indent, but I still want the space between the bullet and the text.

Why would I want this? Lots of reasons. I may have a list that is contained in a table in a page column where it makes no sense to indent the list.

It seems like this should be easily controlled with CSS, but not so. All my attempts at this resulted in results other than what I wanted. Attempts to reduce the left indent usually resulted in the bullet simply vanishing or other odd stuff. I tried having the list items without the actual list tag, which worked in some browsers but not others; I encountered problems with the text wrapping under the bullet.

I thought I must be missing something so I searched a bit online. It seems there really is no good way to do this. Apparently the indented list was intended to be just that - an indented list. Thus, there is no standard reliable way to change it.

What you end up having to do is create a table where each row is a list item with two cells. One cell contains an image for your bullet, the other contains your text. It works, but it's hardly efficient.

Just one of those HTML oddities, I suppose.

Cross-Browser Testing Made Easy

One of the challenges of building a web site is ensuring that it will function the same way when viewed by many different web browsers and versions of browsers on different operating systems. Something that works fine in the latest version of IE or FireFox may not work the same way (or at all) in an older browser or on a different O/S. This is not as much of an issue if you are just using very vanilla HTML. However, if you start using JavaScript, DHTML, or some of the more exotic CSS, you may be in for a nasty surprise when you try to view the page on a different system.

What really got me thinking about this recently was my interest in weaning myself away from table-based design in favor of more CSS-driven design. I know it's supposed to be so much better, the wave of the future and all that. But my concern has always been how to be sure it would work on older browsers.

I try to keep the latest versions of IE, FireFox, NS Navigator, Opera and now Google Chrome on my local machine for some basic cross-browser testing. But I can't have the other versions or the other O/S configurations to test. I don't even own a Mac at present.

I started thinking it would be cool if there was a software tool out there that would grab a page and render it just like a specified browser, version of browser, etc.

As it happens, I found something even better. This web site (http://www.crossbrowsertesting.com/) allows you to log into any of their many and varied images to test your page. It's free, with some restrictions that I'll get to in a bit. But you can hop on to a Mac, Ubuntu or Windows (98, XP, Vista, etc.) system and try out many different browsers, versions of browsers, etc. It's brilliant. It's not some kind of program that mimics the browser. It's the actual browser. And it's fully functional so you can test all your JavaScript, DHTML, whatever.

There are a few restrictions on the "free" part. You can only stay logged into a session for 5 minutes at a time. But you can launch as many sessions as you like. So for quick tests, it is perfectly adequate. Also, paying customers get preference for access when the site gets busy. If you need to do more complex testing and need more than 5 minutes on an image, you can buy little blocks of time. It's all very well thought-out.

Now, if they could just do something similar to test wireless devices...

Thumbs Up for Sphider

The model club site that I run has been steadily growing. We have been adding lots of great content - articles, tips and photos, lots of photos. Our gallery now contains nearly 5000 of them. I am using a gallery application called Coppermine (PHP front-end, MySQL back-end). It's very nice. Each photo has a title and many have more detailed descriptions.

With all this content I felt that the site really could benefit from having some kind of comprehensive search mechanism. Coppermine has its own search, which works fine, but I wanted a way to search the whole site at a go.

My first thought was to use Google Custom Search, which I had implemented with some success on another site. I was able to implement it on the club site without any trouble. The issue that I had was getting it to re-index in a timely fashion when I made changes. I decided that I wanted a mechanism that gave me more control over the indexing. As I have no budget for the club site, I also wanted something that was free.

I found no shortage of free search engines out there and tried a few. But the problem I kept running into was that the free versions had a limit on the number of pages they would index. The limit was high - usually several thousand - but I kept exceeding it. The reason I kept exceeding it was the photo gallery. Nearly 5000 photos, each of which gets indexed as it's own page, plus gallery sub-area pages, etc., etc. That ends up being a lot of pages to index.

I finally found a search engine that I could run locally on my site that was free and had no page limit - Sphider. Sphider uses PHP and requires a MySQL database to store it's indexes. It is really quite nice. Not only can you re-index at will, you can choose to index just certain parts of your site by setting up "sites" in the admin panel that limit their inclusion to just certain areas. This was especially useful for me because I often want to re-index everything except the gallery, which is pretty time consuming due to the sheer size of it.

It took me a little time to get the filters right for indexing the gallery. I had to keep it from indexing certain ancillary pages that had no business showing up in a search result. But Sphider has some decent include/exclude filtering mechanisms to facilitate that. It also respects any directives in your robots text file.

It provides some nice statistics on what search terms your visitors are entering, most popular searches and so on.

Implementation was fairly easy. It uses a template with a header, footer, etc., which gives you enough flexibility to make it a seamless part of your site. Once your MySQL database is in place, you just pass Spider's admin panel your db user credentials and it takes it from there.

All in all, really not bad. I have had it in place for about 2 months now and it seems to work really well.

New Window or Same Window?

Any time you create a link on a page, you have the option to launch the target page in the same window (the default option) or in a new window. We aren't talking about those obnoxious automatic pop-ups here. We are talking about links the visitor actually clicks on. There seem to be several schools of thought on when it is appropriate to launch a new window.

Some say you should never launch a link in a new window for a couple of reasons:

• The user can always launch in a new window by choice by holding CTRL when they click the link or by middle-clicking the link with the mouse wheel.
  
  
• By contrast, there is no easy way to make a link coded to launch in a new window launch in the same window. So, you are essentially taking the choice away from the visitor.

I'm not sure I agree with that. For one thing, many users don't know how to launch in a new window. Also, in many cases, a user may want to check out something you are linking to without leaving your site. I like the following rule of thumb, which seems to be the consensus from what I have read:

• Make internal links (links to stuff on your site) open in the same window.
  
  
• Make external links (links to other sites) open in a different window.

There are some exceptions to the first point. For example, if you are displaying a short form or a Flash piece or something and you want better control over the window like size or toolbar/no toolbar, etc. But for the most part I would think these rules would work.

I'd be curious if anyone agrees/disagrees either from a designer standpoint or from a user's standpoint. Let me know...

Fun With Google Maps API

I was a Geography major in college and I have always loved maps. I like looking at them and I like working with them. It is something I often have to do when I build web sites for small businesses, particularly those with physical locations that customers need to get to.

In the past, I always just put together a static map graphic and added a hyperlink to MapQuest or some other site for directions. Then I read about Google allowing people to tap into their maps API to display fully functional Google maps on their own pages. (I read about it in Quest Software's Knowledge Xpert for MySQL, of all things, the latest version of which includes a very nice tutorial on this subject.) Anyway, it's really pretty easy. Just a matter of signing up with Google to get a key and then adding some JavaScript to your page - not that dissimilar to Google Analytics, another wonderful and inexplicably free Google web toy.

To start, you need to register and get a key from Google. You need one key for each domain you want to put your map on. You can get a key here:

http://code.google.com/apis/maps/signup.html

Google will provide you with the code you need to drop into your page and they have loads of docs on the various options available to you. Alternatively, you can get some nice code from the aforementioned Knowledge Xpert product, which is free.

But it's straightforward enough. Some JavaScript in your HEAD, then a DIV tag where you want to actually place the map. You can include or exclude all the various controls like pan, zoom, Map/Satellite/Hybrid toggles, etc. It can be easily sized.

One thing that's not as obvious. Some of the Knowledge Xpert examples center the map using a Google geocode (the lat and long) for where you want the map to center on initially. Their instructions for obtaining that geocode for a specific address were a little fuzzy for me. As it happens, there is a URL that you can ping to get the code for any address. The example below finds the geocode for 810 Guadalupe Street, Austin (if you put in your google maps key at the end):

http://maps.google.com/maps/geo?q=810+Guadalupe+Street,+Austin,+TX&output=csv&key=your_google_key_here

To see the example where I used Google Maps, visit http://www.austinsms.org/meetings.php and have a peek at the source.

Knowledge Xpert also provides additional information on tying this into a MySQL database to store and map multiple locations, if you are so inclined.

Lots of fun to play with and a much nicer solution to providing a map on your web site.