This comment by Lassi HippelÃ¤inen deserves notice:
Sorry if I sound pedantic - I worked many years as a system architect in computer secutiry - but this argument will not go anywhere, unless its basic terminology is clear. More specifically, there are two concepts that are getting mixed up all the time: anonymity and pseudonymity.
Pseudonyms are stable, used by the same person(s) all the time. They have Internet credibility, even if the real name is not publicly known. Therefore even pseudonymous writers tend to behave civilly. "Anonymous" bloggers are in fact pseudonymous.
Anonyms are unstable. They can be created and disposed at a whim. Next time they might be used by some other persons. Anonymous commenters don't have to worry about credibility. They can shoot from the hip, and be fraudulent.
The real problem is anonymity, not pseudonymity. Without anonymity there wouldn't be astroturfing or sock puppetry either.
Even Nisbet is unclear about the difference. Concluding from his text, he only wants to block anonymous, not pseudonymous comments. He accepts a Facebook identity, which from the technical point of view is a pseudonym, because he has no way of confirming that the name belongs to the person who created the account.
Whistleblowing does not require anonymity, only pseudonymity.
From the mouth of Finns!
Anonymity, as opposed to pseudonymity, has its utility in particular circumstances. For my own blogs though I discourage anonymity, and often won't publish comments without distinctive handles. I call them "drive-by comments." Sometimes they're useful, but a disproportionate number of these sorts of comments are not useful, and are made without understanding the context of the post or discussion. If I ran a "muck-racking" website where "tips" were useful or encouraged, I'd probably have a different attitude.
Of course, even pseudonymity is often superficial. I've looked up many commenters via their email address, and often can find a Facebook page with their "real identity." At a minimum IP information is often very informative. For example, a Muslim fundamentalist who was jabbering about how decadent and disgusting the West was naturally was posting his comments from Sheffield, UK. Another time I noted that a commenter exclaiming how incredible a multicultural society was actually was resident of a very homogeneous city, and that their EU citizenship allowed them to make different choices if they so chose. Revealed preferences matter. Another time I was having a discussion with a commenter who occasionally posted under a pseudonym, and the issue were assumed aspects of my personal life (the assumptions were by and large false). Since my own biography came up, I decided to look up the individual's Facebook page, and then told them that I now knew their real name, associations, and had seen their photograph. I explained that now I felt more comfortable having a discussion on an equal playing field when it came to our personal experiences, but the commenter simply disappeared. I inferred from that the commenter was comfortable about talking about these issues with personal valence only if their own identity was something they could construct on their own terms, where they had all the power to shape perception and I naturally had less (since aspects of my own background are relatively well known).
Finally, over my nearly 8 years of blogging I have found that the odds ratio of insulting and "non-constructive" comments is greater from anonymous than from pseudonymous comments, and from pseudonymous than "real name" comments (again, using email addresses I've find that a non-trivial number of commenters use names that might not be their real name). But naturally that doesn't mean that only commenters who use their real name are useful; most of the really informative and edifying commenters are pseudonymous, because a disproportionate number of comments are pseudonymous. Pseudonymous comments often tend to have more verve and punch than comments from people using their real names, so it isn't as if there is a spectrum with only one dimension. On occasion because of the nature of this weblog I also receive comments from anonymous "drive-by" commenters which are extremely detailed, and exhibit a level of fluency with the material which warrants serious consideration. When I do an IP trace I often find that likely the commenter was somehow associated with, or a critic of, the research which I'm putting the spotlight on. The reason for anonymity here is rather obvious. Finally, sometimes when a discussion is going on I will intervene and note that pseudonymous commenter X "has creds" after doing a quick background check. Obviously I'm not going to disclose the specific information, but that fact should help reorient the discussion in a direction which is to everyone's advantage.*
The blogosphere is then I believe an information ecology where different levels of personal disclosure may be optimal in serving everyone's needs. The level of optimal disclosure will naturally vary by blog and the nature of the content and comment. I will, for example, tolerate a relatively personal critique from someone who uses their real name and email address which I likely would not if they were a pseudonym, and certainly would not if they were anonymous. Parity of disclosure is important if you're going to operate in the zone of interpersonal familiarity. Finally, with subtle moderation technologies often the downsides of anonymous commenters in terms of incivility can be dampened. Actually demanding an email address, even if fake, actually weeds out 90% of the trolls.
* I have never disclosed specific personal information which could be identifying when I've found an anon or pseudo's identity, but, when an anon or pseudo is extremely abusive or unproductive, disclosing that I actually know who they are invariably results in their disappearance. Though on one occasion I noticed the person reappeared later with a different pseudo, and their comments were much more constructive.
The level of anonymity appears to affect the mindset of commenters. Even in whistle blowing cases though, the anonymous tipster must establish credibility with the individual receiving the tip. (One reason why ability to protect sources is so important in the scheme of journalism.)
So while anonymous commenters might potentially have something of value to add, they forget that somehow they must establish credibility for that information to be taken seriously. Without going via the pseudonym route this is very difficult. (And of course it takes all the fun out of spewing.)
As you mentioned - maintaining true anonymity in a commenting session is very difficult. I'm hoping that as folks learn the ropes of the new world of electronic communication it actually forces an improvement in the level of dialog - since our words invariably catch up to our identity if we bother to say anything that matters.
I use a pseudonym when I discuss things on blogs, because I would like to remain gainfully employed and I don't want to try to guess what my employer might think of my opinions every time I post something.
Indeed, my pseudonym is stable and not used to hide behind for snarky attacks.
I not only use pseudonyms when I comment on blogs, I use different pseudonyms (or anonymous comments) on different blogs, because I don't want some automated software someday tying together everything I've ever written on the Internet for the benefit of some casual inquirer. I have non-controversial opinions on non-controversial subjects, and opinions that would really offend some people, and at the very least, even if nobody is going to be showing up on my doorstep, I don't particularly want those opinions mixed together on-line.
Having multiple pseudonyms doesn't completely eliminate the risk though. People tend to reuse expressions and arguments that they think are particularly telling or clever, and these can be tracked through Google. Even if you manage to avoid using the exact same wording in different venues, people have distinctive writing styles, and I can imagine that someday there may be software that can search on that.
Back in the day, I published a couple of moderately controversial thoughts under my real name on this newfangled thing called Usenet, secure in the knowledge that whatever I wrote there would disappear forever in a few weeks. Then along comes this other thing called DejaNews, and suddenly my words are preserved for all eternity. (You do realize, don't you, that people could very well be reading these archives a thousand years from now). Thank goodness I had the foresight (um, was sufficiently paranoid) even then to go to the considerable trouble of getting a second email address (which wasn't free back then) for my most transgressive opinions. But if I should ever reuse some some particularly clever argument from that old alias (and I've probably already done so, more than once), in principle I could still get nailed. Yes, truly, we are all headed into a new age of transparency.
To add to prior comments: Keeping one's own identity from being exposed on the Internet is indeed very hard. Even if you use TOR or a similar system to hide your IP address, there are still other ways your IP address can be exposed (most notably via plug-ins), not to mention other techniques like analyzing writing style, cross-correlating cookies and other web traces across multiple sites, and so on.
My general philosophy has been to always use my real name when commenting on blogs that are related to my work or that address topics that I've blogged about myself, to avoid the appearance of sockpuppetry. I use a pseudonym only on other sites like GNXP where any opinions I might express are casual at best, given that I don't have any relevant experience in the field and haven't thought and written that deeply on the topic. But even then I'm aware that the veil of pseudonymity could easily be pierced, and measure my words appropriately.
I think the key factor involved may be reputation. Almost by definition, an Anonym has no prior reputation to ride on, and no future reputation to aspire to.
A Pseudonym's reputation may build to a level equal or superior to a True Name (which in one sense is just another type of Pseudonym); may approach the zero (or perhaps, zero-point) of an Anonym; or may acquire a dis-reputation. Where disreputation is not the deliberate objective (such as to tarnish the reputation of others by association) and on recognizing that significant disreputation has developed, a Pseudonym user has the opportunity for effecting an instant gain in reputation by abandoning the old pseudonym for a new one. Conjecturally, this will happen as soon as the perceived cost of losing the old pseudonym (which may have sentimental value as a "cool name") and the personal cost of developing a new pseudonym (which may involve a deliberate cooling-off time delay before comments are allowed, may involve e-mail registration of the claim to the alias, or may just require thinking up a new alias) are outweighed by the gain in reputation. In the case of a True Name, the pseudonym has very high costs for repudiation and replacement, in some cases (usually for disreputation resulting from action, rather than speach) risking criminal sanction.
Such churn on internet pseudonyms has two easily identified social costs. Long term, there is a finite number of pseudonyms for any length; a form of "address space" exhaustion can result (and be hastened if useful contributor John6254 is posting alongside notorious troll John6245). However, this is not yet an major problem, save perhaps for the Smith-s and Li-s of the world. More immediately significant, if troll turn-over is high enough, the base level of reputation for arbitrary new pseudonyms is lowered; all newcomers start in a deep social hole, limiting the ability of the network to continue growing. The evolutionary response has been for social networks to provide tools for reputation tracking; EG, Slashdot comment points, Fark favorite/ignore settings, Ebay buyer/seller feedback, and others. These have design trade-offs with effectiveness, meta-reputation, and ease-of-use.
Such reputation may be a factor for why "Pseudonymous comments often tend to have more verve and punch than comments from people using their real names", in that there is less existing reputation at risk when using a pseudonym; people are more willing to take chances when the potential loss is smaller.
Possibly interestingly related: (doi:10.1145/355112.355122), (doi:10.1145/775152.775242), (doi:10.1007/s10458-005-6825-4). More recent work has been done, but my quick poke at Google Scholar doesn't turn up anything that provides a more definitive mathematical model.
There's also the potential for reputation to be a better considered as a vector rather than a scalar. (EG, someone who has an excellent reputation on pure mathematics, but whose commentary on politics is widely considered utterly bonkers.)
I have a real name, and I have a nom-de-net. You want to know my real name, you need but ask. I have no qualms at letting people know who I am, but I won't condemn others for hiding behind a nom-de-plume for safety's sake.
Hiding your name behind a pseudonym doesn't go far. There are already now automatic tools that analyse text and help in identifying he author. They are used, e.g. to find plagiarism. That is the main reason why I have never bothered to develop a pseudonym. I'll get "outed" anyway, sooner or later.
BTW, the first such tools were developed already in ninth century Arabia to analyse the texts in Koran. As a side product they lead to the first known cryptographic crack, frequency analysis.
Anonymity also has a non-symmetric aspect. As Razib points out, the blog owner has far more details about the commenters than the readers. Someone may be anonymous to us, but pseudonymous (or even known) to him. That raises the issue of lending credence (i.e. tacitly endorsing) comments that pass moderation.