Roadrunner newsgroups - up down down downup down up down down down down
Ragnar Suomi
unstable?
It's so annoying. Why can't they fix it? They obviously don't know
what the problem is.
Maarten Andriessen
"Ragnar Suomi" <ragna...@yahoo.com> wrote in message
news:5ht68099bg843fiiq...@4ax.com...
RoadRunner will never fix it. What part of that is it that you do not
understand?
Maarten
Mike Yetto
> On Mon, 19 Apr 2004 12:28:07 GMT, Leythos <vo...@nowhere.com>
> scribbled:
>
>> it without using the crap about investing money into it. Try
>> explaining
>
> You mean I'm supposed to explain the reason without using the
> reason why?
>
That is what the original RFCs intended.
>> give you a hint, money is a part of the problem, but how about
>> addressing the technical side of the problem.
>
> Beyond the theoretical "The feed is just too big" its ALL about
> money.
>
Why not just tell him that it is no more worth Roadrunner's effort to
keep up with the ever increasing feed than it is to explain anything
to him.
Mike "just what nourishment do trolls get, anyway?" Yetto
--
"There is danger from all men. The only maxim of a free
government ought to be to trust no man living with power to
endanger the public liberty."
- John Adams
Mike Yetto
> Well, he's supposed to never be replying to me again.....
He doesn't want to violate your Fifth Amendment right to due process.
Mike "isn't excessive whinge covered under the eighth?" Yetto
Maarten Andriessen
> 1) Usenet without the ALT.BINARY groups would be 99.9999% complete and
> up all the time based on most ISP's current infrastructures.
>
> 2) There is no ROI on upgrading servers to provide content that is other
> than text for usenet groups.
>
> 3) Bandwidth of all users is impacted by users that pull binary objects
> from Local usenet servers - this decreases the amount of bandwidth
> available to all local users. 100 text only groups can be pulled from
> local servers in about 20 seconds (headers and bodies), but binaries can
> take hours of continuous streaming data across the network.
>
> 4) Storage space - Binaries eat a LOT of storage space - there is no ROI
> for it.
>
> 5) Inbound PIPE Size - in order to pull a single days feed, based on the
> ever increasing feed, you need a massive (read that as expensive) feed
> just for the news server function. The same feed/pipe to carry the text
> groups amounts to about 8% of that volume.
>
> 6) Usenet is a terribly inefficient method of moving files.
>
> So, if you take into account all of the above REAL ISSUES you can
> clearly see why they don't invest a lot of money into Usenet services -
> there is NO ROI to carry the binaries.
Add to that the following:
7) RoadRunner consists of frnachises that each have their own investments.
Each franchise has its own newsservers, and it s own limited budget to
operate them. Nationwide ISP's like Earthlink have 2 larger newsserver
farms, one in the east, one in the west. Because of the franchise structure
of RoadRunner the local frnachises have to fend for themselves.
Taking point 1 through 6 in account, you can see that it is a very
complicated and EXPENSIVE business to run newservers properly.
There is a reason why newsproviders started with charging $4.95 when they
started, then $9.95 for a while, and I think most of them are now charging
by the gigabyte instead of a fixed rate. Well, you have "package rates" like
$5.95 for 3 Gb and $10.95 for 10 Gb.
Reason 6 is the best reason why it simply does not work for binaries. I
think if they removed the ILLEGAL binaries from the servers it would work a
lot better though.
Maarten
WyleC...@cactus.com
In other words ....... things are normal
Maarten Andriessen
> Can you provide a link that refers to this RR "franchise" structure?
> AFAIK, RR *is* a national ISP, with service delivery primarily via
> TWC, it's parent company's cable infrastructure. There are a couple
> other contracts in place I believe, such as with Brighthouse Cable,
> although that may only be a transitional deal.
Yes it is a national ISP. The reason why I *strongly suspect* it is a
franchise (I have heard it mentioned in forums on dslreports too but cannot
find it) is the completely different pricing the business subscriptions have
(even the residential ones vary) and the different rules that each office
operates. On top of that, the websites of all the local time warner
"franchises" are completetely different!
For instance, some areas allow you to bring your own modem, others don't.
Some areas use completely different brand modems then others.
Some areas install different equipment brands then others.
Other areas charge you $5 more if you don't take on cable TV, others do not.
Some areas (like in Western Ohio) have 3072/512 while most others do not.
There are so many differences between the different offices, that it either
HAS to be a franchise structure, or the national ISP simply has no control
over their own people. The local websites have no consistency whatsoever,
different color schemes, different setup, different writing style, etc.....
> There is a TWC/RR regional structure, but as it's wholly owned by RR
> and/or TWC, it doesn't meet any definition of a "franchise" that I'm
> aware of...
Maybe it is not a "franchise" structure, but the regional offices have a lot
of influence on the way things are handles. Including usenet.
RR Could have a much better usenet system if they were to throw their 30
news-servers on 2 big heaps and make a big farm.
Maarten
MTV
> In article <5ht68099bg843fiiq...@4ax.com>,
> ragna...@yahoo.com says...
> What part of all of these threads have you not been reading.
>
> If you follow the text only groups you will find that they are very
> stable and have a good retention time - this does vary by the regional
> location, but for the most part they are doing very well.
>
> If you are looking to usenet to download files, you will need to look
> somewhere else, or subscribe to a service that is designed to handle the
> warez and porn that makes up the vast majority of the files passed
> around the world in Usenet servers.
>
And don't let the door slam on your way out!
Binaries should be banned from RR.
MV
Ragnar Suomi
The question is, why aren't they working? Some days they work
perfectly and most others they don't.
Also, it is not just affecting binary files. It's only noticable in
binary files becasue they are thousands of times larger than text
messages. The odds of parts missing in a binary are much greater, and
you see the partial binary.
With text messages, they are almost all one part. If the part is
missing, you never even know it existed.
It's as if a malicious attack is being done to roadrunner. The
partial files are typically 95 to 99% complete. There must be someone
intentionally causing random parts to not propogate onto the servers
to make them practically unusable.
On Wed, 21 Apr 2004 22:14:37 GMT, MTV <NoSpam...@houston.rr.com>
wrote:
Ragnar Suomi
If a text message is incomplete, you never know it because it's only
one part typically.
The percentage of missing parts is the same whether text or binary.
You notice it with binaries because they are a set of hundreds of
parts and the binary itself is often thousands of times larger than a
text message.
The question isn't whether or not these binaries should be carried.
The question is why is it that they work sometimes and don't work most
of the time?
On Mon, 19 Apr 2004 11:18:47 GMT, Leythos <vo...@nowhere.com> wrote:
>In article <5ht68099bg843fiiq...@4ax.com>,
>ragna...@yahoo.com says...
>What part of all of these threads have you not been reading.
>
>If you follow the text only groups you will find that they are very
>stable and have a good retention time - this does vary by the regional
>location, but for the most part they are doing very well.
>
>If you are looking to usenet to download files, you will need to look
>somewhere else, or subscribe to a service that is designed to handle the
>warez and porn that makes up the vast majority of the files passed
>around the world in Usenet servers.
>
>--
Tim S. Knight
server.columbus.rr.com:
>
> Maybe you can explain how a RR server is going to pull all 500 parts of
> a single file when one or more parts expires on the source server before
> the pull is complete?
>
Any real news server handling a 'full' binary feed isn't going to any
pulling at all. The articles will be pushed to them by the servers they
peer with.
Dave Kelsen
> While still snuggled in a 'spider hole', Leythos <vo...@nowhere.com> scribbled:
>
>>So, Sun servers, used all across the country, by more than just RR, have
>>problems that limit their ability to provide service to applications
>>running on them at RANDOM times?
>
> RR apparently bought into some new system Sun was offering, as a test.
> Needless to say, the test wasn't all that good.
> New HD arrays went in on the Orange County news server(s) yesterday,
> so things are better, but the feed still sucks, and we're up to maybe 60%
> complete on binaries.
>
>>If you have Sun's document(s) on known problems, please post a link to
>>them and where they address problems with Usenet servers for all of us
>>to see.
>
> I'm relying on what the Head Admin here in San Diego is telling us.
> I've never known him to lie to customers.
Dude, you know an *Administrator* for head? Awesome!
--
RFT!!!
Dave Kelsen
Dying is no excuse. Nixon in 2004.
Tim S. Knight
news:MPG.1af22072e...@news-server.columbus.rr.com:
> In article <Xns94D3BBE...@corp-radius.supernews.com>,
> spu_...@yahoo.com says...
> Nice of you to qualify that - but fact is that servers can get a list
> and then do a pull or they can be pushed to. I have now idea which
> method is used in our region.
>
For any sizeable usenet feed it would be a CHECK/TAKETHIS stream. Pulling
articles off a server as a reader would creat too much load on the server.
> And to put your "push" into perspective, if the source push has all
> the parts, but the destination server expires ANY part of the article
> before the push completes you end up with incompletes - this would be
> due to the amount of data being pumped over the connection and the
> rate that it's being processed.
>
That's why anyone with clue peers with many different servers in order to
make sure that inevitable gaps in the feed.
Mike Yetto
> In article <reqg80tg8a0l6eh88...@4ax.com>,
> niobrara969 @yahoo.com says...
>> On Thu, 22 Apr 2004 23:26:42 GMT, Leythos <vo...@nowhere.com>
>> wrote:
>>
>>> I have now idea
>>
>> Finally...the admission I've been waiting for.
>
> Too bad your are only good at taking things out of context. This
> could have been an interesting exchange, but you've done
> everything you can to be crude/offensive at almost every post.
>
>
Bill, you have to admit; the guy can dance.
Mike "so can Michael Jackson" Yetto
Tim S. Knight
news:MPG.1af22803f...@news-server.columbus.rr.com:
> In article <Xns94D3C04...@corp-radius.supernews.com>,
> spu_...@yahoo.com says...
>> That's why anyone with clue peers with many different servers in
>> order to make sure that inevitable gaps in the feed.
>
> And if the article expires on the local RR server before the remainder
> of the 100 parts get here too, then you get incompletes. You keep
> saying it, just not understanding it.
>
For that to happen it would have to be a something that takes extremely
long to post. However, you weren't talking about posting, so I'll just skip
that.
Text articles move across Usenet in milliseconds, binaries take a few
minutes, but still not long enough that they would expire before all of the
parts got propigated. So you'd have to have a server with extremely short
retention (several *minutes*.)
Tim S. Knight
news:MPG.1af25434...@news-server.columbus.rr.com:
> In article <Xns94D3E18...@corp-radius.supernews.com>,
> spu_...@yahoo.com says...
>> Text articles move across Usenet in milliseconds, binaries take a few
>> minutes, but still not long enough that they would expire before all
>> of the parts got propigated. So you'd have to have a server with
>> extremely short retention (several *minutes*.)
>
> Are you assuming that all the parts of a multi-part post are sent in
> order one after the other? They go in the same order as they were sent
Unless the news server is doing some strange queuing voodoo, posts go out
in the order they were received, with smaller posts moving faster.
> to the source server - meaning that they could be intermixed with
> other posts for the same group. Depending on the system (network)
> capacity it could take an hour for all the parts to make it from the
> source (first person to start posting 100 parts) until it makes it all
> the way to our server with 100 parts - and a part could expire in that
> time.
Yes, there is a chance that it could take an hour. But there is also a
chance that the UPS in the data center could explode and the server could
burn down. It's possible, but not extremely likey unless you are having
network problems, that affect all of your peers. Most news servers have
serveral, and they offer outgoing posts to *all* of them.
Ragnar Suomi
>In article <iusf80hcn1omsgn0r...@4ax.com>,
>ragna...@yahoo.com says...
>> It's as if a malicious attack is being done to roadrunner. The
>> partial files are typically 95 to 99% complete. There must be someone
>> intentionally causing random parts to not propogate onto the servers
>> to make them practically unusable.
>
>There is not attack, no conspiracy, no mystery, it's just how usenet
>works when there is SOOOOOO much data being pulled for multi-part posts.
>It's been this way for as long as people have been posting binaries to
>usenet.
>
>In the old days, before binaries were part of usenet, it was much easier
>to get a full days pull in a couple hours and to retain the entire group
>list on a 5 or 10 gb of space (5 or 10mb in the real old days).
>
>I can assure you that if RR stopped carrying the binary groups, limited
>messages to (random number) 200 lines, that usenet would work perfectly
>for about 95% of all ISP's out there.
>
>--
It has nothing to do with retention.