This is probably relevant if you have a blog or run a webzine of some sort. It'll be much more abstract if you are a commenter, and can't relate concretely to weirdo creeps who persistently spam your comments and contact you via email. In relation to bloggingheads.tv my own two primary complaints from my experience on that web-show:
1) A group of commenters, lead by one particular commenter, persistently verbally abused me as a racist repeatedly even if I was talking about something totally unrelated to race (the commenters admitted that their goal was make sure that I was not invited to the show again through the campaign of harassment). Of course most of these individuals were anonymous. I told the producer that I wasn't excited about this in relation to my future possible appearances, and he admitted that they were looking into correcting these issues.
2) The commenters are not, in my opinion, as smart as Bob Wright thinks they are. One regular commenter would offer opinions on evolution based on what Stephen Jay Gould write. This a problem for anyone whose bread & butter is evolution (I'm sure the commenter is not knowledgeable enough to know that this is a problem, though at least Bob Wright acknowledged this specific issue politely). Another reoccurring tendency is that comments sections should serve as forums where one can correct factual errors on the part of the guests. Factual errors occur in the unedited heat of the moment (read some of my posts, and you see that all the time). But often when I skim the comments I don't see the appropriate correction. On one or two occasions I left a quick comment pointing out obvious problems in a guests' assertions because no one else had. When it comes to quantitative social metrics guests routinely are off by significant multiples, and no one seems to correct them in the discussion thread.
On the more general problem of comments, trolls, etc., one must address the question of norms, values, and ends. If I was blogging to maximize short-term pageviews I would encourage robust discussion of every single post by as many people as possible. I don't encourage robust discussion by as many people possible because I will trade pageviews for quality of discussion. If people comment, they often feel invested in a post, and will come back over and over to engage. But sometimes you don't want people to be invested, because they aren't contributing genuine returns to your "bottom line." Of course, if you want pageviews, you are getting returns back to the bottom line. But if you want informative and intelligent discussion which will elevate more than the egos of specific and limited interlocutors then you aren't adding anything. Over the long-term even maximizing pageviews by encouraging commenting can be counterproductive, as is addressed in the discussion. Once comments turn into trash it is very hard to reverse that perception, and people don't read them, aside from the regular trolls for whom it is a hobby to foul up the public spaces of the internet.
My own empirical assessment has been well articulated repeatedly:
1) Most people have very little worth saying on most topics.
2) Some people have a great deal worth saying on some topics.
To encourage the latter to comment and take time out of their days and contribute to the public circulation of information you need to squelch the former. At different points individuals may be in class #1 or class #2. When was the last time I wrote about Hellenistic influences on Umayyad art? I've read a little bit about this topic, but I don't have anything worth saying.
One particular aspect of writing on the internet, and encouraging on-point comments, is that people contribute comments in relation to how they perceive the writer. A major reason I repeatedly identify as a conservative on this weblog is that for several years there was a strange tendency of liberal self-congratulation in the comments, with the mistaken assumption that I was a liberal who would laugh along. Since I blog about science this is a reasonable prior, so I periodically attempt to "update" readers. I think it works. The only problem is that sometimes the new model is that I'm as stupid as other conservatives when it comes to the details of cultural diversity and history, when the reality is that on any given topic of cultural diversity and history you should bet on me, and not anyone else, as knowing what they are talking about (i.e., Roger Bigod may know more about American colonial history, with a focus on Virginia, Paul Conroy knows far more about Ireland than I do, and Patrick Wyman has forgotten more about Late Antiquity than I have, but I am willing to bet that I know more about the other domains outside of their specialty than any of them. I would enjoy meeting someone who knows as much about Roman history and Chinese history as I do, but I haven't met that person yet. Feel free to step up and engage me simultaneously on both).
Which brings us back to norms, values, etc. My main goal as a blogger is to get paid in insight and information. To maintain this stream requires a careful balancing of openness to new interactions, but a strong discrimination against superficial and glib entrants into the public forum. And this being a social system the individuals themselves are not static. I have been introduced to books by readers, and have introduced books to readers. We are in some ways objects of emulation when it comes to particular techniques and sources of personal self-cultivation (I have acknowledged that Dave Munger and Zachary Latif have influenced my own production a great deal, as I have emulated aspects of their own styles; one could add other bloggers, readers, and scholars who I have met via the blog).
All of this requires an almost ecological and organismic mindset. I obviously aim for the highest quality and quantity of readership. But I understand that those who comment must be a narrowly constrained subset of the readership. And even among the commenters there are the casual transients, and the regulars. And finally among the regulars one must assess their strengths and peculiar domains of specialty. This world is most definitely not flat. Rather, this is an information pyramid, with very few apex organisms as one ascends up the cascade of info-trophic layers.