I've used both types of platforms, often with differing goals in mind.
Creating my own space enables me to provide or define a structure into which contributions are placed. I've found the structures (or structurelessness) of virtually all third-party platforms to be ultimately highly distracting and unsuited.
(My own structures also usually show their own weaknesses over time. This doesn't keep me from trying to improve on them, and perhaps find a different form (e.g., non-hiearachical, multiple dimensions or facets) that affords greater utility.)
Organising and managing all of this is incredibly time-intensive. (The fact that I've been more-or-less attempting to do this under a set of increasingly adversarial external circumstances for much of a decade ... rather proves this point.)
If a personally-managed platform gives structure, then large public fora give reach, exposure, and increased potential for engagement (though very often of a much lower quality than a specifically-focused space). These are trade-offs. I make them with a strong awareness of them.
Virtually any general-public forum becomes immensely noisy. Whist illuminating tangents are gold, irrelevant or very tired ones are rather less so (see HN's own guidelines for avoiding shallow and/or dogmatic diversions). Having to repeatedly address the same very basic points becomes tiresome, and the inability to bundle, aggregate, and address in mass repetitive and tiresome points is a particular source of pain. Temporality is very often the principle (or only) organising mechanism, and any substantive discussion developing over time is actively deprecated and pushed out of public view by newer material, that almost always of vastly lower interest. (Power laws: high value is of much lower frequency.) Time-ordered presentation inherently promotes low-quality content.
(This affects HN as well. I've commented that at least via Algolia Search it's possible to find the highest-ranked (if not necessarily best) stories of a given period: day, week, month, year, or other: https://news.ycombinator.com/item?id=28806795)
An extreme frustration I have with federated protocols to date has been their lack of effective search. Mastodon and Diaspora* both exhibit this. HN has been more useful to me in many ways due to its excellent Algolia-based search capabilities. Reddit has no comment search at all, but pretty powerful post-based search, which for "self-post" based subreddits proves reasonably useful.
I've remarked for years on how the instructional method used by the Scholastics of lectio, meditatio, and quaestio seems as if it might be very useful to resurrect, though how to do this effectively in an asynchronous distributed format is a real challenge. Paraphrasing slightly:
- lectio is Reading the Fine Article. It emerged as practice because books were literally worth more than gold, and too expensive to grant students their own copy. A university lecture was often literally the professor reading the book to the class.
- meditatio was an interval in which the content was considered and reflected on, before questions and disputes were raised.
- quaestio was a submission of questions based on the passage. Thse were asked, but not immediately answered (a practice still observed by some institutions, the London School of Economics comes to mind where 5--6 questions are asked, then answered in one pass by the speaker). Disputationes were explorations of controversies.
It's all very ... mediaeval ... but also at least ensures that the disucssion 1) is based on having been exposed to (if not informed by) the material and 2) questions and commentary are moderated and managed in a way that should encourage substance and minimise disruption.
https://en.wikipedia.org/wiki/Scholasticism#Scholastic_metho...
What we have now faces markedly different circumstances:
- Reproduction of content (textual, audio, images, video) is trivial. Getting everyone to access that at the same time is ... harder. (Oddly: in the broadcast world, especially before personal recording systems were prevalent, this was less the case.)
- Conversations occur with participants distributed across space, and often at least staggered in time. Most significantly focus and attention are rarely concentrated on any one specific discussion. Spatial orientation, literally gathering participants within a single chamber, affords far greater attentive orientation. We are trading reach for focus.
- Assuring and assessing RTFMedness is difficult. Much discussion is initiated by little or no awareness of the actual material at all. (Mind: I do this myself as much as anyone, though I try to be aware of it and chide myself not to.)
- Questions and discussions tend to be less rather than more moderated. There are exceptions, but they're rare. Odds that a given discussion remains productive and reasonably focused on a large / general-access site are quite low. Where threads can be personally-managed (e.g., Diaspora*), I've managed to at least have a few smaller-scale, longer-term, and productive discussions (often 4-10 participants, but lasting over days or weeks, occasionally months).
All of this of course assumes that substantive discussion (or even just good shitposting) is a primary goal of such platforms. Most often it's not; advertising reveue is the overarching consideration. Sadly this is sort of an anti-Midas touch: everything it comes in contact with turns to shit.
It's also long struck me that pairing a discussion and wiki (or other distillative format) site would be a superpower of sorts. Reddit's exceedingly anaemic Wiki functionality offers the merest of hints of this potential power, and the fact that that's not been more significantly developed remains a major disappointment.