[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Pan-users] Scoring troubles...
From: |
Duncan |
Subject: |
Re: [Pan-users] Scoring troubles... |
Date: |
Mon, 4 Aug 2014 01:45:50 +0000 (UTC) |
User-agent: |
Pan/0.140 (Chocolate Salty Balls; GIT d447f7c /m/p/portage/src/egit-src/pan2) |
JB posted on Sun, 03 Aug 2014 10:40:47 -0500 as excerpted:
> Hi all,
>
> I've been getting a large number of the german-netherland big MB-sized
> posts to a certain newsgroup. Unfortunately because I can only afford
> the cheapest option of the server for my groups,
This doesn't directly address your question (see below for that), but it
/might/ solve the "cheapest option" problem for you. Perhaps not, too,
at least not immediately, but it could be worth considering for later, if
you for instance get a lump sum (like a tax refund) once a year or some
such.
Here's the deal. While most NSPs do month-to-month, there's a few that
do unexpiring block downloads. If you don't do a whole lot of huge
binaries (and it sounds like you simply can't, ATM) and/or if you do
binaries but your downloading tends to be in fits and spurts (my case, I
can go a year or more without touching them, then have a bit of time and
get the download bug and be rather active for a week or two, then months
to over a year without, again), these unexpiring block accounts tend to
save you **LOTS** of money over even the cheapest month-to-month account.
OTOH, if you're quite consistently downloading GiBs a day, month-to-month
accounts tend to be a better option. But the breakover is higher than
you might think...
As an example, astraweb.com (see below for another vendor with similar
block account plans), where I have my current block account, works like
this. First, their gigabyte counts are 10-power, but they do discount
headers and only count 20% of them, so 10 MB of headers means you get
charged for 2 MB of download. With that in mind:
Block size Once-off Price (USD) = Per-gig price
25 GB $10 = 40 cents/GB
180 GB $25 = 13.9 c/GB
1000 GB $50 = 5 c/GB
Now I don't know what your monthly plan allows or its cost but do the
math. Take even the 25 GB plan at $10. Suppose you download a GiB a
day, which is about what I might do when I'm active. That's still 25
days worth of downloading and a lot of people are paying $10/mo. Suppose
you have a really cheap 3 months for $10. That's still a quarter gig a
day, and I can't imagine you getting three months for $10 and being
allowed a quarter gig a day of downloads
And that's just the smallest block tier and thus the most expensive per
gig available. A $25, 180 gig block very likely compares quite well to
your yearly cost.
What I did here is spring for the 1000 GB block for $50, a bit over a
year ago. I figured at my usage, that may well last a lifetime, and I
was right. In a bit over a year I've used just under 2 GB of it. At
that rate, I have 499 more years worth. Now my usage is as I said in
fits and starts and I actually used it only a few days, but even if I
started doing in a week what I did in over a year, that's still nine and
a half years more worth of service.
Even if I doubled /that/, it'd still be ~5 years worth of service. $10/
year? I could live with that. =:^) But there really is a good chance
I'll literally never have to worry about buying a news account again, for
the rest of my life, and I basically don't have to worry about how much I
use it.
Tho I know someone that downloads several entire TV series as they come
out, buying hard drives for a few tens of dollars apiece to store them
on, one series per drive, and giving them away as presents to friends and
family, a year's worth of their favorite series every Christmas or
whatever. He says he does tens of GB per day in news downloading. For
him, an uncapped monthly account (astraweb has them for $15/month speed
uncapped, or $10/month, limited to 10 Mbit/sec.) probably does make more
sense.
But it doesn't sound like you're anywhere near that. If your news
traffic is what it sounds like, then those block accounts really are
likely to both save you money and reduce or eliminate your severe
downloading limit worries.
If you like the idea of block accounts but don't like astraweb for some
reason, or just want another option to compare it against, blocknews.net
offers similar unexpiring block accounts. They actually have many more
block account tiers, as small as a 5 GB block for $2.75 (55c/gig), and
are cheaper for the smaller blocks (25 gig for $8.50, 34c/gig, vs. $10 at
astraweb, 40c/gig), but don't drop as rapidly. They only offer 500 gig
at the $50 price-point (actually $51.49, 10.3c/gig), with their biggest
block 3072 gig for $240 (7.8+c/gig, still higher than astraweb's $50 for
1000 gig, 5c/gig). Until a couple years ago, however, astraweb's biggest
block was the $25 180-gig, and when astraweb first came out with the 1000
gig it was $100 or 10c/gig, and blocknews was undercutting astraweb at
that point, tho I think astraweb is better known and has a better
reputation, at least in the US.
There may be other unexpiring block providers as well, but those are the
two I know about. And I haven't had any complaints about astraweb.
Except... AFAIK astraweb is US-based (tho they have a European server
near Amsterdam as well, actually I think most serious NSPs do as that
seems to be a major news hub, I believe due to somewhat more liberal
censorship/copywrite/content laws than most western nations, US and
European alike), while I believe blocknews is European based, altho they
have a US server too. If you're concerned about the NSA and recent
Snowden revelations, etc, blocknews may well be a better choice for you.
Tho for all I know (which is saying I don't), they may /both/ be NSA
bait. <shrug> Tho a government subsidy due to to it being the bait for
an intelligence project /would/ explain how they can offer block accounts
undercutting everyone else like that and still stay in business. OTOH,
if their raw per-gig costs are low enough that those charges are covering
them, then interest on the unused balance could explain how they manage
the unexpiring bit, too. Either way, there's enough there to feed a
conspiracy theory if that's your bent, but that's pretty much always the
case for those with that bent, so that's not actually saying much.
Just saying, the block account option is there, and it's worth
considering if you're stuck at the low end. If your budget is tight, you
may have to wait and scrimp a bit to come up with the lump-sum for the
block, but then again, blocknews' low-end 5GB/$2.75 block isn't hugely
costly, and it's just possible that one alone is a better deal than your
presumably monthly current NSP service.
Anyway... on to answer your question, now...
> just *one* of these
> danged posts makes it so I don't get to see any other posts. For
> example, when I open that group, I normally do a 'download latest 300
> headers'. What I get when I do that now is *one* header and it's just
> a spam-like thing and the 'who' that sends it constantly changes so I
> can't make a score on the sender.
Keep in mind that due to the fact that pan combines headers for multi-
part posts, such that many individual pieces of the post and thus
individual message headers, may be combined into one. If it's say a 1000-
piece multi-part and you download 300 headers, you may indeed get 300
headers all from the same multi-part post, which pan will combine into a
single header. If it's not complete, pan will add a (300/1000) or
similar to the end (it might be [300/1000], I forgot which way they
usually mark segments vs. parts), but either way, that *one* header pan
displays may actually be *many* individual headers, combined into one for
display (this is one of the ways pan reduces memory usage, allowing it to
handle millions of headers where it used to crash at a few hundred K
headers).
The point being, looks can be deceiving. That *one* header as displayed
may actually be rather more than that, if pan detects that they are all
parts of the same multi-part post.
> One guy said to make a score of 'fx*.fr*' on the headers, but Pan
> doesn't have a way to do that.
>
> I see a 'references' that brings up this for the spammer:
> <address@hidden> (and each one is
> similar but *always* different).
FWIW, the @ sign in that triggered gmane's address munging algorithm, so
those of us who read this list in pan via gmane's list2news service won't
see the original references-ID. If you think it's worth the hassle, you
can avoid such munging by separating the @ symbol from the content on
either side using a space.
> I've tried adding a scoring to 'references, matches regex' and putting
> <fx.*.fr*> but this has done absolutely nothing so I know I'm doing it
> incorrectly or Pan isn't working right (more likely the former, heh
> heh).
>
> Can anyone tell me how to score this a-hole to -9999 so that I can
> start getting the regular messages in this group, please?
Well, if the given ID wasn't munged, above, and/or if I was familiar with
this particular situation... chances are I probably could.
What I can point out is this: If you check the pan-devel list, there's a
very recent thread (ATM the most recent) with SciFi as the OP, that
mentions something very similar to this, which he calls the "alphabet
soup problem". I don't know if it's the same exact sort of spamming
you're seeing or not, and thus whether the scoring method discussed there
will work for you or not, but it's worth a shot...
Here's a link to the first post of that thread on gmane's web interface.
>From there, clicking the topic link on the left should get you the
thread, including my answer (and another) to his scoring question.
http://permalink.gmane.org/gmane.comp.gnome.apps.pan.devel/1560
With a bit of luck you're talking about the same thing and the suggested
filter, entered directly in the scorefile if necessary, will help.
--
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman