Technology Tales

Consolidation

For a while, the Windows computing side of my life has been spread across far too many versions of the pervasive operating systems with the list including 2000 (desktop and server), XP, 2003 Server, Vista and 7; 9x hasn’t been part of my life for what feels like an age. At home, XP has been the mainstay for my Windows computing needs with Vista Home Premium loaded on my Toshiba laptop. The latter variant came in for more use during that period of home computing “homelessness” and, despite a cacophony of complaints from some, it seemed to work well enough. Since the start of the year, 7 has also been in my sights with beta and release candidate instances in virtual machines leaving me impressed enough to go popping the final version onto both the laptop and in a VM on my main PC. Microsoft finally have got around to checking product keys over the net so that meant a licence purchase for each installation using the same downloaded 32-bit ISO image. 7 still is doing well by me so I am beginning to wonder whether having an XP VM is becoming pointless. The reason for that train of thought is that 7 is becoming the only version that I really need for anything that takes me into the world of Windows.

Work is a different matter with a recent move away from Windows 2000 to Vista heavily reducing my exposure to the venerable old stager (businesses usually take longer to migrate and any good IT manager usually delays any migration by a year anyway). 2000 is sufficiently outmoded by now that even my brother was considering a move to 7 for his work because of al the Office 2007 files that have been coming his way. He may be no technical user but the bad press gained by Vista hasn’t passed him by so a certain wariness is understandable. Saying that, my experiences with Vista haven’t been unpleasant and it always worked well on the laptop and the same also can be said for its corporate desktop counterpart. Much of the noise centered around issues of hardware and software compatibility and that certainly is apparent at work with my having some creases left to straighten.

With all of this general forward heaving, you might think that IE6 would be shuffling its mortal coil by now but a recent check on visitor statistics for this website places it at about 13% share, tantalisingly close to oblivion but still too large to ignore it completely. All in all, it is lingering like that earlier blight of web design, Netscape 4.x. If I was planning a big change to the site design, setting up a Win2K VM would be in order not to completely put off those labouring with the old curmudgeon. For smaller changes, the temptation is not to bother checking but that is questionable when XP is set to live on for a while yet. That came with IE6 and there must be users labouring with the old curmudgeon and that’s ironic with IE8 being available for SP2 since its original launch a while back. Where all this is leading me is towards the idea of waiting for IE6 share to decrease further before tackling any major site changes. After all, I can wait with the general downward trend in market share; there has to be a point when its awkwardness makes it no longer viable to support the thing. That would be a happy day.

Out of memory at line: 56

This is an error that I have started to see a lot in the last few weeks. First, it was with Piwik and latterly with WordPress.com Stats. For the record, I have never seen it on up to date systems but always with IE6 and at page unloading time. The CPU usage hits 100% before the error is produced and that has had me blaming JavaScript in error; it isn’t the cause of all ills. In fact, the cause seems to be a bug in a certain release of Adobe Flash 9 but I am of the opinion that the inclusion of certain features in a Flash movie are needed to trigger it too. I don’t have the exact details of this but WordPress.com Stats worked without fault until a recent update and that is what is making me reach the conclusion that I have. That observation is making me wonder whether we are coming to a point where Flash compatibility is something that needs to factored into the use of the said technology in a website or web application. Updating Flash will solve the problem on the client but it might be better if it wasn’t triggered on the server side either.

Self-hosted web analytics tracking

It amazes me now to think how little tracking I used to do on my various web “experiments” only a few short years ago. However, there was a time when a mere web counter, perhaps displayed on web pages themselves, was enough to yield some level of satisfaction, or dissatisfaction in many a case. Things have come a long way since then and we now seem to have analytics packages all around us. In fact, we don’t even have to dig into our pockets to get our hands on the means to peruse this sort of information either.

At this point, I need to admit that I am known to make use of a few simultaneously but thoughts about reducing their number are coming to mind but there’ll be more on that later. Given that this site is hosted using WordPress software, it should come as no surprise that Automattic’s own plugin has been set into action to see how things are going. The main focus is on the total number of visits by day, week and month with a breakdown showing what pages are doing well as well as an indication of how people came to the site and what links they followed while there. Don’t go expecting details of your visitors like the software that they are using and the country where they are accessing the site with this minimalist option and satisfaction should head your way.

There is next to no way of discussing the subject of website analytics without mentioning Google’s comprehensive offering in the area. You have to admit that it’s comprehensive with perhaps the only bugbear being the lack of live tracking. That need has been addressed very effectively by Woopra, even if its WordPress plugin will not work with IE6. Otherwise, you need the desktop application (being written in Java, it’s a cross-platform affair and I have had it going in both Windows and Linux) but that works well too. Apart maybe from the lack of campaigns, Woopra supplies as good as all of the information that its main competitor provides. It certainly doe what I would need from it.

However, while they can be free as in beer, there are a some costs associated with using using external services like Google Analytics and Woopra. Their means of tracking your web pages for you is by executing a piece of JavaScript that needs to be added to every page. If you have everything set to use a common header or footer page, that shouldn’t be too laborious and there are plugins for publishing platforms like WordPress too. This way of working means that if anyone has JavaScript disabled or decides not to enable JavaScript for the requisite hosts while using the NoScript extension with Firefox, then your numbers are scuppered. Saying that, the same concerns probably any JavaScript code that you may want to execute but there’s another cost again: the calls to external websites can, even with the best attention in the world, slow down the loading of your own pages. Not only is additional JavaScript being run but there also is the latency caused by servers having to communicate across the web.

A self-hosted analytics package would avoid the latter and I found one recently through Lifehacker. Amazingly, it has been around for a while and I hadn’t known about it but I can’t say that I was actively looking for it either. Piwik, formerly known as PHPMyVisites, is the name of my discovery and it seems not too immature either. In fact, I’d venture that it does next to everything that Google Analytics does. While I’d prefer that it used PHP, JavaScript is its means of tracking web pages too. Nevertheless, page loading is still faster than with Google Analytics and/or Woopra and Firefox/NoScript users would only have to allow JavaScript for one site too. If you have had experience with installing PHP/MySQL powered publishing platforms like WordPress, Textpattern and such like, then putting Piwik in place is no ordeal. You may find yourself changing folder access but uploading of the required files, the specification of database credentials and adding an administration user is all fairly standard stuff. I have the thing tracking this edifice as well as my outdoor activities (hillwalking/cycling/photography) web presence and I cannot say that I have any complaints so we’ll see how it goes from here.

Suffering from neglect?

There have been several recorded instances of Google acquiring something and then not developing it to its full potential. FeedBurner is yet another acquisition where this sort of thing has been suspected. Changeovers by monolithic edict and lack of responsiveness from support fora are the sorts of things that breed resentment in some that share opinions on the web. Within the last month, I found that my FeedBurner feeds were not being updated as they should have been and it would not accept a new blog feed when I tried adding it. The result of both these was that I got to deactivating the FeedBurner FeedSmith plugin to take FeedBurner out of the way for my feed subscribers; those regulars on my hillwalking blog were greeted by a splurge of activity following something of a hiatus. There are alternatives such as RapidFeed and Pheedo but I will stay away from the likes of these for a little while and take advantage of the newly added FeedStats plugin to keep tabs on how many come to see the feeds. The downside to this is that IE6 users will see the pure XML rather than a version with a more friendly formatting.

position: static?

CSS positioning seems to be becoming a nightmare when it comes to IE6 support. While I am aware that the likes of 37signals have stopped making their products work with it, there remain a lot of people who stick or are stuck with the old retainer. I am one of the latter because of the continued use of Windows 2000 at my place of work, though a Windows Vista roll-out has been mooted for a while now. If nothing else, it keeps me in the loop for any inconsistencies that afflict the display of my websites. Positioning of an element within the browser window rather than within its parent element is one of these and it looks as if specifying a position of relative in a stylesheet is part of this. Apparently, it could be down to its non-triggering of IE’s haslayout property. It might be a hack but I have found that static positioning has helped. I’ll continue to keep my eye out for a better solution if it exists but the static option seems to have no detrimental effect in IE7, IE8, Firefox, Safari, Chrome or Opera.

Running Internet Explorer on Linux

On first sight, this probably sounds daft given how good Firefox is but you cannot ignore those surfing the web using the ever pervasive Internet Explorer when doing some web development. Using virtualisation is a solution to the need but it can mean that you need to set up a web server with Perl, PHP, MySQL and the like in a virtual machine, all for a little offline testing and then there’s the potential for a lot of file copying too. Otherwise, you are trying to sneak things online and catch the glitches before anyone else does, never a good plan.

Therefore, having the ability to run IE to test your offline LAMPP set up is a boon and IES4Linux allows you to do what’s really needed. Naturally, WINE is involved, so some flakiness may be experienced, even after the ever useful API library’s reaching version 1. Otherwise, all usually runs well once you work your way through the very helpful instructions on the IES4Linux website. I did get a misplaced message about the version of WINE that I was using, and Python errors made a worrying appearance, but neither compromised the end result: a working IE6 installation on my main Ubuntu box.

IE5 and IE5.5 are also on offer if you’re interested but, after looking at my visitor statistics, I think that I can discount these. IE7 and the work-in-progress IE8 make no appearance on the availability list. The absence of IE7 is not a big problem as it might appear because coding for IE6 sufficiently suffices for IE7, even now; IE8 may not be the same in this regard but we shall see. Even so, a later browser release does mean a more secure version and I reckon that including IE7 should be next on the project’s to-do list. Saying that, what we have now is far better than nothing at all.

JavaScript: write it yourself or use a library?

I must admit that I have never been a great fan of JavaScript. For one thing, its need to interact with browser objects places you at the mercy of the purveyors of such pieces of software. Debugging is another fine art that can seem opaque to the the uninitiated since the amount and quality of the logging is determined an interpreter that isn’t provided by the language’s overseers. All in all, it seems to present a steep and obstacle-strewn learning curve to newcomers. As it happens, I have always found server side scripting languages like PHP and Perl to be more to my taste and I have no aversion at all to writing SQL.

In the late 1990’s when I was still using free web hosting, JavaScript probably was the best option for my then new online photo gallery. Whatever was the truth, it certainly was the way that I went. Learning Java or Flash might have been useful but I never managed to devote sufficient time to the task so JavaScript turned out to be the way forward until I got a taste of server side scripting. Moving to paid hosting allowed for that to develop and the JavaScript option took a back seat.

Based on my experience of the browser wars and working with JavaScript throughout their existence, I was more than a little surprised at the buzz surrounding AJAX. Ploughing part of the way through WROX’s Beginning AJAX did nothing to sell the technology to me; it came across as a very dry jargon-blighted read. Nevertheless, I do see the advantages of web applications being as responsive as their desktop equivalents but AJAX doesn’t always guarantee this; as someone that has seen such applications crawling on IE6, I can certainly vouch for this. In fact, I suspect that may be behind the appearance of technologies such as AIR and Silverlight so JavaScript may get usurped yet again, just like my move to a photo gallery powered on the server side.

Even with these concerns, using JavaScript to add a spot more interactivity is never a bad thing even if it can be overdone, hence the speed problems that I have witnessed. In fact, I have been known to use DOM scripting but I need to have the use in mind before I can experiment with a technology; I cannot do it the other way around. Nevertheless, I am keen to see what JavaScript libraries such as jQuery and Prototype might have to offer (both have been used in WordPress). I have happened on their respective websites so they might make good places to start and who knows where my curiosity might take me?

Missing borders in Internet Explorer

It’s quite hard to describe this observation in a title so here goes with a longer description in a post. One thing that I spotted with the Prosumer theme used on this blog is that the links on the horizontal navigation bar underneath the mast head were not appearing as they should. The links have been formatted using CSS to appear in boxes with borders that are more apparent when you hover over them. In IE, the top and bottom borders were missing. After a spot of digging, I came up with the line-height property being the cause and I was right: the extremities of the boxes surrounding the text were being cut off because they exceeded the allotted space. As if to emphasise that IE7 isn’t as major a leap forward from IE6 as we would have liked, the problem affected that browser as well.

Aside: Link text colours weren’t being honoured by IE7 like they are by IE6, Firefox and Opera so another tweak to the CSS was needed.

An inappropriate use of JavaScript

I have seen a web application that displays thousands of records in a scrollable table (please bear with me, there is a very good reason for this). from the appearance of the table, it would be reasonable to assume that the table is generated by the server and output directly to the screen but this isn’t the case. What actually happens is that the server more or less outputs JavaScript code that is then executed. This takes the form of large arrays that are slotted into the DOM as the contents of the required table by a JavaScript function. With the large amounts of data involved, this means that the browser fully loads the client CPU while the JavaScript processing takes place, something that takes up to a minute to complete. Admittedly, the browser is IE6 but this was all on a PC with a 2.53 GHz Pentium 4 and 512 MB of memory. Getting the server to deliver standards-compliant (X)HTML for what is needed in the first place seems a much, much better approach to me.

IE6 and JavaScript performance

Having been exposed to an application at work that uses a lot of JavaScript, I fully appreciate what some mean when they talk about IE6’s inefficient handling of JavaScript. After seeing a web page taking an age to reload and your CPU taking a hammering because of JavaScript processing, the penny does tend to drop… Needless to say, this very much impacts the world of AJAX-driven web applications with their heavy dependence of client-side JavaScript. While IE7 does come to the rescue, there remain plenty of IE6 users still out there and this is reflect in website statistics. This reflects a certain level of inertia in the browser market that not only afflicts the uptake of IE7 but also the likes of Mozilla, Opera and Safari. It also means that anyone developing AJAX applications very much needs to continue testing in IE6, especially if the product of their labours is for wider public use. An example of such an application is Zimbra, an open source web application for messaging and collaboration, and the people behind it have generously share the results of their browser performance benchmarking. They did comparisons of IE6 vs. IE7 and Firefox 2 vs. IE7. IE6 easily came out as the worst in these while Firefox 2 was the best. I suppose that the next question to be asked centres around the type of code that is processed inefficiently by IE6. I wouldn’t be at all surprised if a list emerged but here’s one: using Microsoft’s proprietary innerHTML object to update the DOM for a web page format. Having a quick trawl on Google, this came up for mention as a cause of memory leaks. It is also a Microsoft innovation that never got taken up by those overseeing web standards, hardly a surprise since a spot of DOM scripting achieves the same end. It may be faster to code than any alternatives, and it does have some support from other browsers, but it does seem to have got a bad name and so should be avoided if possible. That said, it would be interesting to see a performance comparsion between innerHTML and DOM methods in IE6.