My server has 30,000 newsgroups. You only report 200.
Welcome to the world of statistics! It is not necessary to measure every individual to obtain accurate averages of an entire population.
There are MANY more newsgroups on the servers than are sampled. Active files are multi-thousands on most servers, and only about 200 are sampled. Only the newsgroups in common to both servers make up the report.
If you are doing poorly in the sample 200 newsgroups, you are VERY UNLIKELY to be doing better in any others.
As long as the servers involved treat articles in the reported newsgroups the same as in the 30,000 newsgroups, the reported performance is valid and represents the performance in all similar newsgroups. See also: Can an estimate be made for newsgroups not sampled?
There are newsgroups which are not "similar" to the sample newsgroups, (I.e. they are not treated the same.) Some newsgroups have a very large percentage of articles which are automatically cancelled or rejected by spam or other filters. Some examples of this are:
- Newsgroups with very large postings, sometimes not propagated (warez, multimedia)
- ECP and EMP, especially in alt.binaries.pictures*, or the *jobs* groups
- "rogue cancels" in controversial newsgroups.
- non-binaries newsgroups which have a lot of binaries posted
- newsgroups not widely carried (bogus names, superseded, etc.)
- newsgroups marked moderated on some servers, but not others.
These types of newsgroups are not in the sample set. In practice, these newsgroups vary widely from server to server, and would be very difficult to report reliably. (Reporting performance in these types of newsgroups may be done in separate part of the newsrAte report in the future.)
For more information see Why sample these 200 newsgroups, and not some other set?
Cheating?
It is possible for sites to "stack" the sample newsgroups by getting a selective feed. This is akin to tuning a system to perform well on industry standard benchmarks at the expense of general performance. There are ways to detect "stacking" and this technology should make its way into the sample software eventually. The sample set size may also be increased, to make stacking more difficult.
The primary purpose of newsrAte is to provide accurate estimates to news server administrators. It is not intended to compare two competing news services. It is possible that reports you see published include "stacked" performance enhancement, so judge wisely.
View an example report
Can an estimate be made for newsgroups not sampled?
Why sample these 200 newsgroups, and not some other set?
What if some sample newsgroups are not on the server?
How do I signup?
Up to The newsrAte RKT
Up to newsrAte home
Up to Mib Software home
Copyright 1998, Forrest J. Cavalier III, Mib Software
INN customization and consulting
Email: newsrate@mibsoftware.com
Voice: 570-992-8824