In traditional computer science, we used to say that
- Hardware lives a few years; then you upgrade to a faster computer
- Software lives for decades: even when you get a faster computer, you want it to keep running your old software. Thus, many companies still depend on software that was written twenty years or more ago, when people thought that the Year 2000 was so far away that it was acceptable to encode dates without recording the century (e.g., 7/4/75). Also, even when software is upgraded, much of the old code lives on in the new version (MS Word users still suffer from bad design decisions made in the 1980s before Microsoft got a usability lab). This is why it's so important to follow the usability guidelines for application design when developing software. Mistakes will haunt you for decades.
- Data lives forever. Once you record, say, a customer's address, you want to keep that information even when you do get so tired of your old hardware and software that you implement an entirely new solution from scratch.
The same is true on the Web. Hardware definitely lives dangerously: any successful site will need to upgrade servers several times a year. And we all know how browsers and other software are in a constant state of flux.
Web data (mainly in the form of pages) should live much longer than Web hardware and software. Even though most users go to the newer pages, older pages will still be of interest to some users. For example, at Sun we still have customers using almost every product we ever shipped, so information about these old products is still of interest. Even the sales pages will be of interest to any third-party customer who might be thinking about buying old equipment from a company that has moved to fresh machines. And it is in our interest to support these third-party customers even if we don't make a cent from them buying used equipment: they may take out a service contract, and they will certainly become prime prospects for buying their own upgrades at a later date.
Consider, as another example, a user who is thinking about seeing the 1946 Humphrey Bogart film The Big Sleep . Sure, he or she can find a modern review in, say, Cinemania , but wouldn't it be more interesting to see what The New York Times wrote about the film back in 1946? Certainly, film students would want to know how the film was received under the circumstances for which it was produced. This example shows that the Times would have a better website if they had 50-years old pages online. Similarly, the review of this year's new Evita film could easily be attracting thousands of hits in the year 2020.
The conclusion is clear: pages designed today may well be used many years from now, so designers are advised to mark up the information as close to the standard as possible. Also, try to create information with persistent value as far as possible. Sure, you could always go back and fix up old pages (just as people have to hire expensive consultants to solve the "Year 2000" problem in their software because they didn't think far enough ahead originally), but it will be expensive and the likely outcome is that the old pages get discarded (and with them, the opportunities from providing customers with added benefits).