You know the moment. A download runs to 96 %, slows down, stalls. The repair stage kicks in, chews through your CPU for a few minutes, and then your newsreader reports that the post is incomplete. Nothing you can do, no retry that helps.
That is almost never your internet connection. It is almost always completeness — and it is the one specification that separates a Usenet provider that works from one that merely advertises.
What is completeness?
When something is posted to Usenet, it is not stored as one file. It is chopped into hundreds — often thousands — of small, separate messages called articles. Your newsreader collects them all and reassembles the original file at your end.
Completeness (also called completion) is the share of those articles that are actually present and readable on a provider’s servers. It is measured per article, not per file. That distinction is the whole story: a 99 % figure does not mean 99 % of your downloads work. It means that for every hundred pieces your newsreader asks for, on average one is not there.
The trick with the numbers
Every provider publishes a completeness figure somewhere between 95 % and 100 %. Written down, those numbers look almost interchangeable. Drawn as bars, they look identical — which is exactly why the figure is such a comfortable thing to advertise.
Same four numbers, two ways of drawing them. Percentages compress the difference; missing articles expose it. Against a 99 % service we lose five times fewer articles. Against a 95 % service, twenty-five times fewer.
This is the honest way to read a completeness spec: do not look at the number, look at the gap to 100. That gap is what your newsreader has to work around, and it is where the difference between providers lives.
Why repair data is not a free pass
Most posts come with par2 repair blocks — extra data that can reconstruct missing pieces. This is genuinely clever, and it is why a small amount of loss usually goes unnoticed. It is also why some providers treat completeness as a number nobody will check.
But repair has limits, and they matter:
- It has a hard ceiling. A post typically carries around 10 % recovery data. Lose more than the repair set covers and the download is dead — there is no second attempt that fixes it.
- It costs you download volume and time. Repair means fetching extra articles and then rebuilding the file. On a low-completeness provider you pay that tax on almost every post.
- The repair files are articles too. This is the part people miss: par2 blocks live on the same servers, subject to the same completeness. A provider that loses your data also loses your safety net.
- Losses are not evenly spread. Missing articles cluster — around takedowns, around a particular upload, around a moment when a feed hiccuped. An average of “1 in 100” can mean forty consecutive parts gone from the post you actually wanted.
At 99,8% completeness, repair does what it was designed for: it covers the rare gap. At 95 %, it becomes the thing standing between you and a broken download on a regular basis.
Why articles go missing in the first place
No provider is at 100 %, and anyone claiming otherwise should be read with suspicion. Articles disappear for reasons that are partly outside anyone’s control:
- Takedown requests. Specific articles are removed on legal notice. This is the largest single cause and it affects every provider.
- Propagation gaps. Usenet is a network of servers passing articles to each other. If a feed stalls, articles posted in that window can arrive incomplete — or not at all.
- Incomplete uploads. Sometimes the post itself was never complete. Nothing on the provider side can fix that.
- Storage and hardware faults. Rare, but real, and only detectable if somebody is actually watching.
Which is the point: completeness is not a fixed property you either have or lack. It is the result of continuous work — monitoring feeds, comparing against peers, filling gaps, replacing hardware before it fails quietly.
Why the figure holds
Because 99,8% is not something a provider can talk its way to. It comes out of a platform built for exactly that: continuous monitoring, several independent feeds so a gap at one source can be filled from another, and enough storage that nothing has to be thrown away to make room.
That is the platform we sell access to, and we hold ourselves to the number it delivers. If completeness ever slips, it is our problem to chase — you have one address to write to, and we do the rest.
- 99,8% completeness — two missing articles per thousand, measured per article.
- Continuous monitoring, so a dip is noticed on a dashboard rather than in your download queue.
- Several independent feeds, so what is missing at one source can come from another.
- 100.000+ newsgroups carried in full — the quiet ones are not quietly dropped to save storage.
- Twenty years of selling Usenet access under the same company, from the Netherlands.
Completeness and retention belong together
These two numbers are usually listed side by side, and they only make sense together. Retention decides how far back you can reach; completeness decides what condition things are in when you get there. A third number sits next to them: speed and connections decide how quickly those articles arrive once they are found.
Deep retention with weak completeness gives you an enormous archive full of broken posts — you can see the file, you simply cannot finish it. Excellent completeness with a short retention window gives you flawless downloads of nothing but the last few weeks. You need both, which is why we carry 4.000+ days at 99,8%.
See it for yourself
Completeness is the one spec you cannot verify from a website — including this one. So do not take our word for it: grab a few older NZBs, point them at us, and watch what happens in the repair stage.
Pick a plan that fits your usage — every one of them gets the same 99,8% completeness, the same 4.000+ days of retention and the same speed. And if we do not convince you, cancelling takes a few clicks.
