Good though the web is, there are times when the only way to get what you want is to pay for it. You can look at the online versions of various British newspapers, but they only go back a few years. If you want anything earlier, especially if it's American, and you will discover that the archives are only available if you pay for them.
The biggest of the archives systems that I use is lexis-nexis. This is a vast library of material, some of which is only available here. Unfortunately, it is expensive: certainly beyond the reach of any individual freelancer. If you look around on the site, you will find you can purchase individual articles by opening an account and paying by credit-card.
Recently I tried to do exactly that: only to find that the site resolutely refused to accept the existence of any address outside the United States. Bizarre: but if lexis-nexis don't want our money, other people are happy to take it.
One of the best search engines I know is NorthernLight. One of the special things it does is offer paid-for material as part of its normal searches. And it will take your credit card details, wherever you happen to live. Single articles are individually priced in the $1-3 range. Typically, the thing I wanted was from the Wall Street Journal, which likes to make serious money from its archive, and cost me $2.95. There seemed no obvious way of buying it direct from the Wall Street Journal page. One advantage of buying from NorthernLight is the no-quibble guarantee. If the item doesn't serve your "information needs", you get your money back.
Otherwise, you will very often find when you are looking through a newspaper's archives that its holdings are administered by NewsLibrary. And once again, NewsLibrary will happily take your money. Finding those archives in the first place is not too difficult. You can use NewsLibrary's site, which groups them by region, or you can go to this list of online newspaper archives, maintained by Ibiblio, a vast librarian-edited site that calls itself "the public's library and digital archive". I'm ashamed to say I've never looked at it before, but there are a huge number of useful links there.
Still, even though we are learning to pay for things, we all occasionally like something for nothing. That was why so many people flocked to Napster. And there is sometimes a legitimate reason for wanting to hear a piece of music once or twice, perhaps as part of your research, without having to buy the whole CD. Sometimes "streaming" audio will do the job: this is a way of playing music from the net without being able to download it and produce pirate copies. Try Friskit Music, which is the first streaming music search engine I've heard. Streaming music is not an entirely satisfactory experience if you are on a dial-up Internet connection -- you can get heartily sick of the word "buffering' -- but it does work. If you are on a broadband connection, no doubt it will all work much more smoothly.
Tip: the first screens you get for any given artist only list very short 10 or 15 second clips of the kind you hear on Amazon.com. If you want anything longer, click on the "duration" heading above that column. Now your list will have much longer pieces. I found one seven minute piece that claimed to be from an as-yet unreleased Steely Dan album: it didn't sound like that, but you never know, do you?
For some reason, there seems to be a great upsurge in news sites and news aggregators (which bring together a lot of news feeds in one place). I have already praised RocketNews, and I find myself using it most days. It faces a lot of competition, however.
AllTheWeb, also known as Fast, now has a strong news search facility. It tells you how long ago the page was collected, which is useful. But it does not list its sources anywhere, and British coverage doesn't seem to extend much beyond The Guardian and the BBC.
Then there's a very ambitious site known as the World News Network. Unlike, say, RocketNews, this one puts all its sources in the shop window and lets you choose. It also has an impressive string of specialist searches arranged by geography and business, once you get beyond the very confusing front page. Don't be tempted to use the menu of countries at bottom right, which includes "Edinburgh" but somehown omits Britain. However, a straightforward search on a British story found material from The Guardian, The BBC, The Independent, The Telegraph and The Times, so it did better than AllTheWeb on that test.
The age of the material found by search engines is a continuing problem. AltaVista ran into trouble this year when it was forced into admitting that many of its indexes, especially the geographical ones like altavista.co.uk, were many months out of date. Two new search engines Teoma and WiseNut have been getting good reviews lately. Teoma uses Google-style software that places the most popular sites at the top of its rankings. WiseNut has a very good look and several clever tricks. But both were examined by Chris Sherman, the leading Internet librarian, and found to be serving up pages that were months old. I looked for some reviews of the Honda Civic, for instance, and almost everything Teoma found related to the old model. WiseNut did little better. Clever software and slick graphics are all well and good, but if any of these people are to get within striking distance of Google they will have to work on the basics: building a big index of pages and keeping them up to date.
Meanwhile, you might like to give metasearch another try. A metasearch engine simply applies your search to lots of other search engines' indexes and then gives you the top results from each. Sadly, early favourites like Dogpile have rather ruined things for themselves by choosing to use search engines that supply mainly paid-for results. Unlike some, I don't think paid search is a good thing. If you are doing a shopping search, it may help to have sites from big-spending corporations near the top. It won't help with the kind of research we need to do.
My favourite metasearch engine has always been Ixquick. Its news search is quite effective. I also like MetaEureka, though not for news. But you would do well to take a look at Ithaki. This is an eccentric, enthusiast-driven site, but I like it a lot. Aside from search and metasearch, it searches web directories and will also find news, newsgroups, images, mp3s and, uniquely, MIDIs. (As any musician will know, these are files that drive synthesizers and computer sound chips.) Its homepage is available in 14 languages and, more to the point, it offers metasearch of geographically-based search engines. For instance, it will search five UK search tools, including the useful Mirago.
Best of all, it adds no paid-for links of its own, and uses none of the growing number of search engines that are built around paid-for links. Instead it uses the best search engines around, including Google and Northern Light. Well worth getting to know: how many other sites do you know with references to Cavafy and Rimbaud on the "about" page? Declaration of interest: I have offered to help Ithaki with their rather eccentric English; I hope that won't take away any of the site's distinctive charm.