Message ID Direct
Message ID Direct
Would you consider having message id direct only check servers for messages from the groups that are attached to them? When I turn this on for Newsfeeds specialty servers (for instance), it pings the mp3 groups for vcd's (and vice-versa) thereby wasting time. If it only checked for messages in the groups present on the server, this would not happen. As it is, I'm still downloading headers from these servers in order to not waste this time.
if servers are fast enough it shouldn't be a problem from the bandwidth or speed point of view.
the bandwidth price of a miss is negligible - maybe 100 bytes.
as to the speed, consider the chain of msg-id direct servers as conveyer, even though a miss take time, the time is negligible comparing to the time for downloading bodies, so there will be enough supply to fill the bandwidth. you can also play with strict/non-strict priorities to do what you need.
the scheme should be simple if we add too many options it will be difficult to understand, one may even forget what combination of options he is using at the moment, i mean if there is no tangible gain there is no need in the option.
the bandwidth price of a miss is negligible - maybe 100 bytes.
as to the speed, consider the chain of msg-id direct servers as conveyer, even though a miss take time, the time is negligible comparing to the time for downloading bodies, so there will be enough supply to fill the bandwidth. you can also play with strict/non-strict priorities to do what you need.
the scheme should be simple if we add too many options it will be difficult to understand, one may even forget what combination of options he is using at the moment, i mean if there is no tangible gain there is no need in the option.
OK, but . . . if you consider that a movie comes in 3 disks, 800 megs each, with rar files of 10 MB each, each of those cut to 5000 lines or about 60 odd sections each, that's 3X80X60 or, er . . . 14,400 pings to each server that absolutely doesn't have the file. That's a lot of pings, considering that 2 image servers, 2 mp3 servers, 2 exe servers, and family (and adult, I think) don't carry those groups. The feature is a great idea, don't take this the wrong way. But once you've downloaded the groups for each server on setup and detached those that don't exist on the server, why check the ones that can't have the messages?
I'm not suggesting a user option, but rather that it be made to work that way by default.
I'm not suggesting a user option, but rather that it be made to work that way by default.
just look at it from a different angle, you have a 300K article to download, say you have 10 message-id direct servers, the price of a miss say 100 bytes - 1K for 10 servers or 0.3% of bandwidth or about nothing. From the speed point of view it doesn't matter since 'not found' articles will be supplied faster than they can be downloaded (e.g. if the time to process a miss is 0.2 seconds every 0.2 second an article will be supplied for download directly irrespective of the number of message-id direct servers). it is all in the worst scenario, when message-id servers are strict and they give all misses (which is not the case since then you don't need such servers), otherwise it will be even better.
I don't download many 300k articles, though, and I'm betting that most usenet junkies are more into the 3X800 Meg movies than they are the 300k files.
Using your math of 0.2 seconds per miss, the program may spend 2880 seconds pinging a server that contains none of the articles requested. That's 44 minutes (it will ping looking for every section of every rar file of every movie). If two servers that don't have the article are checked before it finally latches in on one (or more ) that has it, that's 88 minutes. I can tell you that it takes less time to just get the headers and not use the id direct.
Using your math of 0.2 seconds per miss, the program may spend 2880 seconds pinging a server that contains none of the articles requested. That's 44 minutes (it will ping looking for every section of every rar file of every movie). If two servers that don't have the article are checked before it finally latches in on one (or more ) that has it, that's 88 minutes. I can tell you that it takes less time to just get the headers and not use the id direct.
that is why i am using this setup now and it works great for me aleast
http://www.netwu.com/newspro/phpBB2/viewtopic.php?t=68
i don't think there is any other way to get around the 70 times xx/90 posts
other then the way i have setup newspro or if alex can give us an new keyword box to disable / enable ID-Direct timer
http://www.netwu.com/newspro/phpBB2/viewtopic.php?t=68
i don't think there is any other way to get around the 70 times xx/90 posts
other then the way i have setup newspro or if alex can give us an new keyword box to disable / enable ID-Direct timer
i've just explained above why this point seems to be wrong.
ok i'll try to explain it even in simpler words.
say you have 10 junk totally useless working message-id direct servers.
now you download the 3 cds with msg-id direct enabled. suppose it took 3 hours.
then you disable message-id direct and download them again. what i project it will take the same time.
the computer is not like a single human being.
imagine you have 10 workers, you pay them 1 cent a day, you tell them to do some work, how they do it is their responsibility; the task takes them 1 minute. from the other side you could tell every one of them what exactly to do and then it would take them only 50 seconds, it is stuff like that, you just wouldn't bother.
ok i'll try to explain it even in simpler words.
say you have 10 junk totally useless working message-id direct servers.
now you download the 3 cds with msg-id direct enabled. suppose it took 3 hours.
then you disable message-id direct and download them again. what i project it will take the same time.
the computer is not like a single human being.
imagine you have 10 workers, you pay them 1 cent a day, you tell them to do some work, how they do it is their responsibility; the task takes them 1 minute. from the other side you could tell every one of them what exactly to do and then it would take them only 50 seconds, it is stuff like that, you just wouldn't bother.
For those viewing this thread, Alex and I have agreed that the behavior I'm speaking of only happens when certain servers don't have all available Usenet groups, which is mainly a problem with the Newsfeeds service. Some of you may know they offer access to 19 different servers, but many of them are specialty servers which contain only certain groups. For this service, you probably won't see the full benefit of the msg-id-direct feature unless your bandwidth can be filled by a couple of their servers.
More "normal" service arrangements, where you would have a bulk service for most of your file downloads and a backup premium service for completion and retention, will work just as Alex envisioned when the feature was added. Newfeeds customers, especially those downloading a few gig a day, will probably want to turn the feature off on the specialty servers.
More "normal" service arrangements, where you would have a bulk service for most of your file downloads and a backup premium service for completion and retention, will work just as Alex envisioned when the feature was added. Newfeeds customers, especially those downloading a few gig a day, will probably want to turn the feature off on the specialty servers.
please do not restrict Message ID direct only lo attached servers.
There are servers, which don't have the newsgroup in their list, but they have some the articles.
Example: newscache0.freenet.de to newscache5.freenet.de
If the unction would be restricted only to attached servers, the newsache servers would not be catched.
Oli
There are servers, which don't have the newsgroup in their list, but they have some the articles.
Example: newscache0.freenet.de to newscache5.freenet.de
If the unction would be restricted only to attached servers, the newsache servers would not be catched.
Oli