There are actually three purposes for requiring registration, some or all of which may apply at any given site, and none of which add value for the user.

1. Spam prevention. Which can be achieved in other ways, as with captchas. (Some complain that captchas discriminate against the visually impaired, but that's an implementation issue with many off-the-shelf captcha solutions, not an intrinsic problem with the basic concept. It's solvable.) Obviously only applicable where sites require registration to contribute content. Sites that require registration just to browse some content have at least one of the other two motivations.

2. They get something tangible through forced registration. This is the primary motivation when registration is demanded merely to view material. Rather than be grateful for the traffic and ad eyeballs (and, if the user posts something, the free content to attract more ad eyeballs), they figure to extract even more (and usually end up with less, as they drive most prospective users elsewhere on the 'net). This may be reams of data for their marketing department, or for selling to databroker companies like ChoicePoint to sell on for others to misuse; it may be money directly (hint: if a credit card number must be supplied, it's money, even if they claim it's for something else, like "age verification" or "proof of real-world identity". As if they don't know that little Johnny can easily read one off Daddy's credit card! Why not ask for some other form of ID, not usable to transfer money they say they don't want? Because what they say they don't want and what they do want are the same thing! They may not dare ding the account themselves after saying they won't; so they probably sell the number to third parties, i.e. crooks and identity thieves. If they're later questioned, they "had a laptop with credit card numbers stolen". I suggest then asking why they retained credit card numbers that are supposed to be for "identity verification only" or whatever. If they say they retained the numbers, they're up the creek. If they deny retaining the numbers, the "stolen laptop" story develops a hole you can drive a truck through, and they're sunk. Either way, you have them by the short hairs.)

3. Control. Register, and you are now pseudonymous. Your activity may be tracked, more easily than otherwise, and users being registered makes them more susceptible to coercive tactics, such as threatened access restrictions, bans, or other unequal treatment. Invariably, this potential for coercion is handled in a less than equitable and democratic way, and often in a ham-fisted one. A lot of Web forums suffer from dictatorial control and suppression of dissent by a self-appointed class of "moderators". Others have moderators that behave in a relatively fair and equitable manner, but generally still cannot be held accountable in any reliable way by the membership, which inevitably has chilling effects on the discussion of controversial matters. I would tend to prefer an unmoderated Usenet newsgroup fulla spam to the ham-fisted style of moderation, and even the milder-mannered form makes me nervous. As a rule, there is no "constitution" or "user's bill of rights" in such places to enforce fairness and prevent and/or punish "moderators" who "moderate" someone out of personal animus rather than because they actually spammed or otherwise misbehaved in a serious manner.

Also, as a rule, posting pseudonymously visibly ties all your posts together with a single name or handle of some kind, and people cease to consider each one purely on its own merits and begin to confuse message and messenger. Fallacious reasoning of the "This can't be true, because it's coming from so-and-so and he's a jackass" variety gets encouraged.

Of course, it can sometimes help keep a thread of discussion making sense if sequential posts by the same person are identified as such, but that is possible (merely optional) when anonymous posting is allowed. (They can optionally register, if registration is possible, as it is here, or they can simply sign a series of posts in a common way. If they want to make forgeries detectable, they can even pgp sign a posting, in theory. But they can post without the signature anytime they want to, and nobody can "ban" the signature and take it away from them just for disagreeing with Official Doctrine(tm).)

Also of course, even anonymous posts at most places keep some record of the poster's IP address, but that's a lot less intrusive than a registration hoop-jumpathon to the user, a lot less visible an ID (often not visible to non-admins at all), and a somewhat more fluid one, granting a controversial poster (or, unfortunately, a spammer) more deniability. If they are posting certain things from inside places like China, that can be far more important than you, undoubtedly born and raised in a free country, can likely imagine. Posting via a proxy or from a Web café or similar shared computer(s) in combination with no registration greatly increases practical freedom of speech.

The cost? Spammers and disrupters. Only massive, automated bot posting binges pose a genuine threat and a captcha can keep those out. (It can't stop the load functioning as a denial-of-service attack, but neither can requiring registration.) Individual disruptive posters not using automation can have disruptive posts manually deleted. Futility will deter them far more than registration hoops and bans and other punishments. Futility and not even receiving any attention, not even recognition in the form of a ban, will make them look elsewhere for their entertainment. (Disruptive posters trying to use automation are, like spammers, blocked by captchas.)

Individuals are sometimes harder to classify anyway. Some people are simply controversial: opinionated, often insightful, but often angering at least a minority of other users. If postings by them that are purely inflammatory crap disappear without fanfare while postings of genuine value (whether controversial or not; so long as they are on-topic and not purely for the purpose of trolling) are kept, they will quickly learn what isn't worth bothering typing in. No other reinforcement necessary. (Most such posters don't post any "pure crap"; just posts that generate some significant amount of disagreement and maybe on a hot-button issue. Those posters never last long at moderated Web forums and are the life of the party on unmoderated Usenet newsgroups. They are also essential to a functioning society, and Constitutional guarantees of free speech are principally there to protect exactly such personalities. As is often said, it's unpopular or controversial free speech that most needs protecting, particularly dissent the Guys In Charge(tm) would like to stifle. I'd be concerned even with an arbitrary ability to delete individual posts without some mechanism of due process or review of some kind. Quarantine them from general view and invite twelve randomly-picked jurors to decide whether they are crap, maybe; so as not to give widespread attention to the post in the event that it truly is crap and the decision is to delete; but with due process.)

My recommendation: If a site is truly honest and truly has users' best interests at heart (rather than charging them money, robbing them blind, spamming them, or coercing them to toe some company line), then it will permit anonymous participation and use a captcha to keep spammers and any other automated blitzing away. Captchas are a humans-only stateless, accountless access control -- perfect for maximizing user freedom. The only possible reason to object to their use and insist on registration of any kind instead is therefore if there is an ulterior motive that a stateless, accountless access control can't support. Which means $$$, in one form or another. Directly, via spam, or by suppressing embarrassing dissenting opinions.

Think of the money value in that last item! A Web forum for "freely" discussing and reviewing software can make a quiet deal with Microsoft for lots of moolah in exchange for finding trumped-up excuses to ban or otherwise harass, make feel unwelcome, or generally suppress any bad reviewing of Microsoft software. A Web forum run by a company can suppress any prickly remarks (or worse sticky questions) asked about their products or services.

Requiring registration here at the Gripe Log would be, in light of the above, an extraordinarily bad idea. Infoworld would want to exert some editorial control over comment postings, if they don't already with comment deletions. (Could they have anything to do with "supergenius"? Why was respected poster ekuns' post on one particular topic nuked? It surely wasn't spam or anything else users would consider evil, and it was never explained adequately. Perhaps someone else found it evil, someone obviously in a conflict of interest with the user base and therefore patently the wrong person to make such a decision?)

Now consider the pressure Ed would come under to misuse the registration requirement. Symantec would want to suppress user recommendations of NOD32 and AVG. They might be able to name a price high enough to buy even Ed. Microsoft certainly could.

Perhaps they have already, or at least are trying.