Updating definition of "bogus" groups?

Updating definition of "bogus" groups?

[Eric Abrahamsen]

Hi all,

In bug#60078[1], there's some discussion about Gnus's check-bogus-group
functions and whether they're buggy, or what exactly they're supposed to
do, and eventually what "bogus" actually means.

Near as I can tell, "bogus" once upon a time meant something very
specific in the Usenet world, but that was 20+ years ago, and it's no
longer a term of art. If you look at `gnus-check-bogus-newsgroups', the
20-year-old code that determines if a newsgroup is bogus is:

(unless (or (gnus-active group) ; Active
	    (and (gnus-info-method info)
		 (not (gnus-secondary-method-p
		       (gnus-info-method info))))) ; Foreign
  ;; Found a bogus newsgroup.
  (push group bogus))

So if a group is active, or if it belongs to a method that isn't a
secondary method, then it isn't bogus.

That seems like an odd and arbitrary criteria for weeding out groups.

Based on what I think "bogus" once meant (just that a newsgroup is being
superseded, or otherwise discouraged, and dropped from servers' lists),
it seems like a more reasonable update of the intention would be:

A group is bogus if it belongs to a select method that has a
`gnus-request-list' function, but the group itself doesn't appear in the
list.

This seems like it would be more handy for cleaning up "dangling"
groups.

Does anyone have an opinion on that idea, one way or the other?

Thanks!
Eric


[1]: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=60078

Re: Updating definition of "bogus" groups?

[Alberto Luaces]

> This seems like it would be more handy for cleaning up "dangling"
> groups.
>
> Does anyone have an opinion on that idea, one way or the other?

I would remove the term `bogus` altogether, it's too vague, and replace
it with something more descriptive.  This is independent from the fact
that the behaviour could be more useful, as you describe.

-- 
Alberto

Re: Updating definition of "bogus" groups?

[Emanuel Berg]

Alberto Luaces wrote:

>> This seems like it would be more handy for cleaning up
>> "dangling" groups.
>>
>> Does anyone have an opinion on that idea, one way or
>> the other?
>
> I would remove the term `bogus` altogether, it's too vague,
> and replace it with something more descriptive. This is
> independent from the fact that the behaviour could be more
> useful, as you describe.

The "secondary" term is also problematic as in practice it is
a way of using _several_ methods, and they don't necessarily
relate so any hierarchy what is the primary method and what
isn't doesn't make any sense necessarily.

This will probably remain so, but yes, if you build even more
problematic definitions on top of that, it gets even
more problematic.

But instead of getting confused by terminology, which is
always going to be incorrect to some degree in any applied
endeavor, if we are talking `gnus-check-bogus-newsgroups',
what is that it does that is useful? What to do, or not do for
that matter, should be based on that.

-- 
underground experts united
https://dataswamp.org/~incal

Re: Updating definition of "bogus" groups?

[Russ Allbery]

Emanuel Berg <incal@dataswamp.org> writes:

> But instead of getting confused by terminology, which is always going to
> be incorrect to some degree in any applied endeavor, if we are talking
> `gnus-check-bogus-newsgroups', what is that it does that is useful? What
> to do, or not do for that matter, should be based on that.

I've never been clear on what this function does that I would want to do
and why it's bound to a key.  I've only run it when I pressed b by
mistake, and it's never done anything.  I may just be not using the part
of Gnus where it would be useful?

My experience is that when a newsgroup that I'm reading is removed, it
shows up with a * for the article count, so it's pretty obvious that it's
been deleted and I can either delete the corresponding group in Gnus or
investigate what's going on.  But maybe there are scenarios that involve
hidden groups where it makes sense to explicitly check for these?

-- 
Russ Allbery (eagle@eyrie.org)             <https://www.eyrie.org/~eagle/>

Re: Updating definition of "bogus" groups?

[Emanuel Berg]

Russ Allbery wrote:

> I've never been clear on what this function does that
> I would want to do and why it's bound to a key. I've only
> run it when I pressed b by mistake, and it's never done
> anything. I may just be not using the part of Gnus where it
> would be useful?

The docstring says "Remove bogus newsgroups." which, granted,
sounds pretty useful.

But then I wonder, how do they enter the *Group* buffer in the
first place?

> My experience is that when a newsgroup that I'm reading is
> removed, it shows up with a * for the article count, so it's
> pretty obvious that it's been deleted and I can either
> delete the corresponding group in Gnus or investigate what's
> going on. But maybe there are scenarios that involve hidden
> groups where it makes sense to explicitly check for these?

Maybe!

-- 
underground experts united
https://dataswamp.org/~incal

Re: Updating definition of "bogus" groups?

[Russ Allbery]

Emanuel Berg <incal@dataswamp.org> writes:
> Russ Allbery wrote:

>> I've never been clear on what this function does that I would want to
>> do and why it's bound to a key. I've only run it when I pressed b by
>> mistake, and it's never done anything. I may just be not using the part
>> of Gnus where it would be useful?

> The docstring says "Remove bogus newsgroups." which, granted, sounds
> pretty useful.

Historically on Usenet that means newsgroups that were deleted, so I
assume that's what it was intended to mean.

> But then I wonder, how do they enter the *Group* buffer in the first
> place?

Well, if it's just groups that were deleted, one either was subscribed to
them or for whatever reason had them listed and unsubscribed.

In other news readers historically, it was common to keep a newsrc that
included every group on the server whether one was subscribed or not.  In
that context, checking for bogus groups was looking for groups that you
have listed in your newsrc (possibly unsubscribed) that no longer exist.

I guess Gnus could have the same problem?  It definitely doesn't with my
settings, but I know there are a bunch of settings.

-- 
Russ Allbery (eagle@eyrie.org)             <https://www.eyrie.org/~eagle/>

Re: Updating definition of "bogus" groups?

[Eric Abrahamsen]

Russ Allbery <eagle@eyrie.org> writes:

> Emanuel Berg <incal@dataswamp.org> writes:
>> Russ Allbery wrote:
>
>>> I've never been clear on what this function does that I would want to
>>> do and why it's bound to a key. I've only run it when I pressed b by
>>> mistake, and it's never done anything. I may just be not using the part
>>> of Gnus where it would be useful?
>
>> The docstring says "Remove bogus newsgroups." which, granted, sounds
>> pretty useful.
>
> Historically on Usenet that means newsgroups that were deleted, so I
> assume that's what it was intended to mean.
>
>> But then I wonder, how do they enter the *Group* buffer in the first
>> place?
>
> Well, if it's just groups that were deleted, one either was subscribed to
> them or for whatever reason had them listed and unsubscribed.
>
> In other news readers historically, it was common to keep a newsrc that
> included every group on the server whether one was subscribed or not.  In
> that context, checking for bogus groups was looking for groups that you
> have listed in your newsrc (possibly unsubscribed) that no longer exist.
>
> I guess Gnus could have the same problem?  It definitely doesn't with my
> settings, but I know there are a bunch of settings.

What you mentioned about disappeared groups -- that they show up with a
* for the article count -- is connected to how the bogus check works
now: it's all around whether a group can be activated or not. You get a
* when a group can't be activated, or Gnus otherwise can't figure out
the group's article numbers. And the bogus check is also just asking if
the group is active or not.

So I guess what I'm imagining doesn't have much to do with bogosity,
after all. It starts with the servers, not the groups, and asks for an
explicit group list. So maybe it's just something that can live
alongside the bogus check, instead of insisting on replacing it.