There were a few web interfaces, but apart from the hatred the old Usenet users had for web users, there were technical difficultiers that were never overcome.
The biggest one was probably charsets. Back then it wasn't really possible to detect the proper charset for what the user entered into your text field, as far as I was told.
But it was only so bad because users were unreasonable:
de.*, for example, really frowned upon UTF-8 and demanded Latin-1 or Latin-15. Minimal coding, of course, so if you used the Euro sign you had to use Latin-15, if there was no Euro sign in your posting, you had to use Latin-1.
And there were even extremists who declared everything apart from 7 bit ASCII illegal...
> There were a few web interfaces, but apart from the hatred the old Usenet users had for web users, there were technical difficultiers that were never overcome.
To expand upon this a bit it wasn't all just knee jerk elitism. Some web clients were terrible and did terrible things.
WebTV in particular would add animated gif backgrounds and MIDI music. Since Usenet was a text medium those messages would have the gif and midi encoded as text (UUENCODEd?) and added to the post, with a bunch of raw HTML as well. Other users got huge amounts of HTML and uuencoded "stuff" or they got animated backgrounds and MIDI music - either one was pretty awful.
As alluded to later in the replies here, no discussion website ever tried to be fully decentralized like NNTP was from inception.
It is a super hard problem, and as the declining popularity of NNTP itself illustrates (and the tendency toward mega-centralization on the web like Facebook, too) maybe it was solving the wrong problem for today.
I think the modern centralization tendency is a bad thing.
If I was going to redesign usenet today ...?
1. You want non-repudiability in transactions: a given user owns their posts or cancel messages and can be linked to them. So, use a blockchain-based solution for authentication. (Not bitcoin; a separate non-financialized system.)
2. Instead of local spools consisting of a directory tree and sync'd via flood-fill, implement it as a virtual filesystem, peer-to-peer discovery, locally requested contents are cached, so high-traffic newsgroups will be cached more widely and therefore be faster to load.
3. Anti-spam: a big problem is preventing spammers sprouting new sock-puppet identities to get around already-imposed blocks. Hence non-repudiability. I'm not sure how to go about enforcing limits on sock-puppeting: this is a hard problem. But if the authentication mechanism is blockchain based we might be able to link it to an underlying funding system and thereby impose a monetary cost on posting -- even $0.01 per message should be enough to put a drag-brake on spamming. This was always the weakness of usenet: spammers externalized all their costs.
There doesn't need to be an "interface" for NNTP and especially not a web based one. The NNTP client in the old Mozilla products from the '90s was just fine and threaded messages well.
It's more than that. The design of NNTP is much like POP3 - you download entire threads and messages all at once. It doesn't work all that well on the web compared with say IMAP.
my thesis: the atom of NNTP is the message, the atom of web forums is the thread.
(Commenters on that are mostly old BOFHnetters you may still recognise.)