Open Bug 18776 Opened 25 years ago Updated 2 years ago

Show FAQ for newsgroup

Categories

(MailNews Core :: Backend, enhancement, P3)

Product:
MailNews Core
Component:
Backend
Platform:
x86
Windows NT
Type:
enhancement

Tracking

(Not tracked)

People

(Reporter: fur, Unassigned)

References

Details

(Keywords: helpwanted) 

Since the dawn of time, signal-to-noise ratio has been a problem in unmoderated
newsgroups.  In some groups, e.g. n.p.m.general, this is not so much due to
flame wars as it is the result of one-time messages by ignorant users that post
NN4.x questions, are rebuked and then leave in search of another inappropriate
newsgroup to post in.  Maybe this wouldn't happen so much if it was possible
for naive users to discover a group's scope/charter without guessing it from the
group name or groveling through a lot of messages.

What if, the first time that a user visits a newsgroup, the message pane
displayed some sort of "home page" for the group ?  The home page could contain
the group charter, and links to FAQs, archive sites, etc.  Also, the first time
that a user posts to a group, the newsgroup's "post instructions page" would be
displayed in the message pane.  For example, this page might remind the user
that, "questions concerning Netscape products are not appropriate in this
group".

As far as the mechanism goes for locating and loading these per-newsgroup pages,
I'm assuming that extensions to the NNTP protocol are undesirable.  One
possibility is that these pages are conventional news posts that have
distinguished titles, e.g. "!-READ BEFORE POSTING...".  When scanning NNTP
headers, the client would identify such special messages and
automatically download the body of the message.
Severity: normal → enhancement
Yeah, just thinking about this a couple of days ago.  People just don't read
FAQs.
Assignee: phil → nobody
Summary: [FEATURE] Each newsgroup can have a "home page" → [HELP WANTED] Each newsgroup can have a "home page"
Whiteboard: [HELP WANTED]
Interesting idea. Adding to the [help wanted] list.

In actual implementation, there's all sorts of fun in stuff like: what do you
scan? The XOVER output? Or do you actually download the ARTICLE and scan that
for URLs? What if you find more than one? How often do you rescan (supposing the
FAQ URL changes)?
I was thinking about having a FAQ "server" (set to some reasonable default) that
is basically of the form of a URL, which you append the newsgroup to, for
example:

"http://www.mozilla.org/newsgroups/" as the "server" and you might end up with
"http://www.mozilla.org/newsgroups/netscape.public.mozilla.seamonkey".
I thought of the web server idea earlier, but didn't suggest it because it's not
scalable.  You can't have all newsgroups in the world going to a single web
server.  Even if the web server host name can somehow be derived from the news
server name, e.g. news.mozilla.org ==> newsweb.mozilla.org, you can't ask all
the news server operators to also host web sites.  Even if you could convince
them, you have the problem of syncing all the web sites from a centralized
source, etc.  It's better to rely on the home page content being in the
newsgroup itself and be distributed by the existing loosely-coupled NNTP
infrastructure.

In response to Phil's earlier question, you would scan the XOVER data for
special message titles and, if a specially-titled message is discovered,
download the article.  For simplicity sake, I would suggest displaying the body
of the article as the special page, rather than hunting for links within the
page.
Yes, but a web site doesn't need to correspond to a newsgroup server.  You only
need one on the internet, although it's likely there will be competing ones.

Web servers can also redirect to other servers that handle certain parts of the
hierachy.

If you can have all the newsgroup contents going into a site (deja.com), I think
home pages for them all is quite feasible.
If we would use a homepage server, we could 1. just use a redirect to the
homepage (the url shouldn't change that often and redirect don't produce so
much load) and 2. have different servers with some central registry like dmoz.
The load shouldn't be a problem as home.netscape.com still works :-).

I agree, that using (standard) NNTP would be better. But the problem is how to
determine which posts are "official". As soon as spammer notices after which
keywords we scan, I will start seeing random spamming instead of a charta.
The ftp://rtfm.mit.edu/pub/usenet/ archive contains FAQs for most of the
standard newsgroups, and for many alt groups, organized in directories by group
name.
Thanks gwallace for the link. I read a bit, and if I understood it right,
news.answers is moderated, what would solve the spam problem.

We could just emphazise FAQs (crossposts between news.answers and the current
newsgroup) and make one of it the default selection. The summary would read
"Display per default for newsgroups a FAQ post" then.

Is that better than displaying a homepage (redirect from
"http://newshome.mozilla.org/<newsgroupname>" or so) in the msg pane on opening
the newsgroup?
I was thinking a "Read FAQ" button on the newsgroups window, possibly a
drop-down list if there's more than one of them (rec.martial-arts, for example,
has a FAQ split into 4 parts, plus a "newbie FAQ")
Summary: [HELP WANTED] Each newsgroup can have a "home page" → [HELP WANTED] Show FAQ for newsgroup
Updating summary from "Each newsgroup can have a "home page"" to "Show FAQ for
newsgroup".
If this is at all based on the contents of the messages in the newsgroup, you
can guarantee that spammers will hijack it pretty quickly ... Also, you need to
gracefully degrade where the group is obscure and doesn't have its own entry in
the database.

One option that might be worth considering is to show Google search results for
the group's name. If there is an FAQ for the group, doing a Google search would
be pretty likely to find it. (Of course you'd have to be able to use a search
engine of your choice, instead of just Google, so that Mozilla wasn't accused of
partiality.:-)
Matthew (mpt),
news.answers is moderated, if they can't prevent spam, we can't either. Please
see my comment above.
It should "just work", a search means too much work for the user.

gwallace,
My idea was to display the information about the newsgroup (FAQ or Homepage)
directly after opening the newsgroup. The message pane is a graceful place to do
that. It usually displays posts, so all we'd have to do is select the right
post.

A new flag with sort priority 1 (maybe better than highlighting) is useful, if
more than one "official" FAQ in the newsgroup exists, or for reproduction (the
user selected another post, didn't remember the subject of the FAQ, has pref
"remember last selected msg" enabled (so reopening of the newsgroup doesn't
select the FAQ) and wants to go back to the FAQ - well possible). But I don't
think, an first stage of implementation has to wait for these advanced tree
features.
BTW: Matthew, you can fool search engines as well pretty easy.
Yes, but fooling Google is practically impossible. And as for news.answers, I
maintain a newsgroup FAQ which I've never gotten around to submitting to
news.answers. But a Google search for "alt.ascii-art FAQ" works, as expected.
<shrug/>
Keywords: helpwanted
Summary: [HELP WANTED] Show FAQ for newsgroup → Show FAQ for newsgroup
Whiteboard: [HELP WANTED]
Well, both suggestions have disadvantages. Using news.answers restricts the
search to FAQs which the maintainer has crossposted to news.answers. Using
google restricts the search to FAQs that have been posted on the web, and has a
greater (albeit low) possibility of a false positive.

I would go with news.answers. FAQ maintainers are supposed to crosspost to
news.answers. I also think people would just ignore a list of links to FAQs
generated by a search engine, but would be more likely to read a FAQ that shows
up automatically.

One problem in either case is multipart FAQs, like rec.martial-arts (4 parts
plus a newbie FAQ). Another is multi-group FAQs, such as the "Which classical
newsgroup?" FAQ on rec.music.classical.*, rec.music.theory, etc.
My flag-suggestion above can be generalized to a rating field like some advanced

news readers have. It could be an integer value and have the highest sort

priority by default. FAQs would then have a predefined (but customizable) value,

by default very high. This would sort FAQs at teh beginning, and we would just

have to select the first post.



(Later the field you be extended to a full fledged rating system, i.e. a

filtering system, which doesn't move msgs, but alters their rating value.)


Created bug #33296 about the general rating system I mentioned. 33296 shouldn't

be too hard to implement and would make this bug very easy (essentially a filter

rule, which ships as default), if we don't check the crossposting. Adding

dependency.



If we want to stop smart spammer, we'd have to add a new filtering criteria,

which checks, if the crossposted msg is really distributed in the other group,

i.e. if it was accepted by the moderator or not.
Depends on: 33296
> I would go with news.answers. 

I agree about this.  This is what news.answers is for.  Note also that there
are indeed websites that archive the contents of news.answers.  rtfm.mit.edu
has already been mentioned; there is also faqs.org -- but these things in
theory should be available on the news server, which is probably the better
place to get them -- scan the group in question for anything that is 
crossposted to *.answers and fetch it.  Not perfect, but not bad either.
This doesn't give you the charter as such, but what it gives you is at least
as valuable as the charter for _many_ groups, and it should be easy on the
network.  

> If this is at all based on the contents of the messages in the newsgroup,
> you can guarantee that spammers will hijack it pretty quickly ...

*.answers is moderated.  Yes, in theory it is possible to spam a moderated
newsgroup by forging headers to bypass the moderator, but I've yet to see 
a spammer bother to do so; there are so many unmoderated groups...
Product: MailNews → Core
Filter on "Nobody_NScomTLD_20080620"
QA Contact: lchiang → backend
Product: Core → MailNews Core
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.