bbb120

[Batalov]

Originally Posted by retina

Why is it necessary to start a new topic? I haven't seen anything on this board that disallows responding to an existing topic with pertinent information.

Glad you asked!

Most forums do (or recommend) this by default:

Originally Posted by https://communitybuilding.stackexchange.com/questions/2632/why-do-internet-forums-tend-to-prohibit-responding-to-inactive-threads

(see the link for the full discussion)
A necropost is a message that revives (as in necromancy) an arbitrarily old thread, causing it to appear above newer and more active threads. This practice is generally seen as a breach of etiquette on most forums. Because old threads are not usually locked from further posting, necroposting is common for newer users and in cases where the date of previous posts is not apparent.

As posts are primarly ordered by recent activity, it can be annoying (irrelevant contributions make a post come to the top) when there are no levels (answers vs comments) of contributions. In Stack Exchange only when new questions, new answers, or modifications to any of the two are made, the "post" is brought to the top.

The next one came up just recently - and I know that they are probably also have an attack of robots (even though unlike us, they have a filter in place already - "only users with recent PG work credit can post". We don't have that automatic protection option: MF users are not linked to GIMPS credit.)

Originally Posted by http://www.primegrid.com/forum_thread.php?id=8478

Effective immediately, forum threads will be automatically locked one year after the most recent post. If a thread has been inactive for a year, it's better to start a new thread than to reactivate a moribund thread.

If you need to reference something in the old thread, you can always link to the original post or thread using the
/forum-specific/
tag.

[retina]

Originally Posted by Batalov

Glad you asked!

Most forums do (or recommend) this by default:


The next one came up just recently - and I know that they are probably also have an attack of robots (even though unlike us, they have a filter in place already - "only users with recent PG work credit can post". We don't have that automatic protection option: MF users are not linked to GIMPS credit.)

I don't see those sites a relevant to this one. If someone has relevant information to a thread then just add it to the end. Better there than a new thread with no context. Many old threads still have important/interesting information and background. Abandoning them just because they are "old" is a poor criterion IMO.

Spammers are a separate issue. If we have one then ban them. New topic or old topic, it won't matter to a spammer.

[Batalov]

Originally Posted by retina

I don't see those sites a relevant to this one. ...Spammers are a separate issue. If we have one then ban them. New topic or old topic, it won't matter to a spammer.

You contradict yourself - right there.

"New topic or old topic, " ... the damage has already been done! You (as a user) go in to read some interesting discussions and at the top you see threads at the top:

• Record probable prime found!
  
• The probable primes in Five-or-Bust
  
• Record 73-digit factor is found by user...

and open them - only to find that they were -posted to float up.

Also I have a distinct feeling that you didn't read what I wrote entirely. - https://communitybuilding.stackexcha...active-threads
(see the link for the full discussion)
There you would have found:

I've been a moderator on several pretty active forums, and I can offer a few reasons:

It shifts focus away from new posts.

This is probably the number one reason. Most active users sort posts by date so that the new posts go to the top. This allows users to focus on questions and discussions as they come in. Posting to an old thread "bumps" the old thread to the top, which shifts our focus away from new posts to old posts. That might not be a big deal, except:

Most old posts have been abandoned.

Most threads are from new users, mostly with a specific question. That question is then either answered, or it's not answered and some time goes by and the user abandons the question.

Many "bumping" posts are answers to questions that have either already been answered or have been abandoned. If the OP hasn't visited the forum in a year, why spend time answering it? Why make other people spend time reading your post?

It makes old threads harder to read.

Let's say a thread starts, then is abandoned for a year, then you make a post, some discussion ensues, and then another year goes by. Now I as a reader stumble across the thread. The actual answer that was originally in the thread is now somewhere in the middle of the discussion instead of at the bottom. I now have to figure out when each post was made, when the answer was posted, what happened after that, etc. It would be much better for me if this had been split into two separate threads.

This can trigger discussions on abandoned threads.

Another thing I see very often is that one "bumping" post will cause the thread to rise to the top, which will cause other users to jump in, not realizing that the OP is long gone, and continue a discussion that has long since been abandoned. This creates a bit of a snowball effect as more active users spend time in that thread instead of on actual new conversations.

This might not be an issue if the discussion is actually interesting, but most of the time it's not.

Most bumping posts are low quality.

Maybe your bumping post really does add something interesting, and maybe it does spark an interesting conversation. But 95% of the time it's a drive-by hit from a Google search, either offering a solution to an old problem (see above) or saying "thanks me too!" (notice that Stack Overflow disallows posts like this) or saying "I have the same problem, did you ever fix this?"

Those types of posts might be useful to you as a poster, but they aren't useful to the forum as a whole, and they especially aren't useful to power users who are policing the forum.

It's hard to decide whether you're an exception to the rule.

Considering that 95% of bumping posts are low quality, it can be "hard" or time-consuming to take the time to read your post and the thread you've bumped to decide whether your post belongs there or not, then to reply to you asking you to start your own thread, or to lock the thread, or to take whatever action is now required of us.

Most necromancing posts belong in their own thread.

If you think that your post would be useful, maybe because you're having a similar problem and the posted solutions didn't work for you, then your post belongs in a new thread. Instead of continuing the old discussion, start a new discussion and link to the old one.

This avoids all of the problems above, but allows you to ask follow-up questions or start a new discussion.

Our time is free, but it's not worthless.

Keep in mind that you aren't just talking to the OP. You're talking to everybody on the forum, particularly "power users" who sort by date. They're all going to read your post. They're probably going to re-read the entire thread.

That might not seem like a lot of work, but keep in mind that many of these users are also moderating the forum, deleting spam (so much spam), banning rude users, answering questions, etc. We do this for free, in our spare time. You might think "what's the big deal if I make you spend a few minutes re-reading a thread", but moderating a forum is a zero-sum game. I only have a certain amount of time to do all this stuff. The more time you make me spend on old threads is less time I can spend banning spammers. So the more time you take away from us by bumping old threads, the less time we have to actually do our "jobs" on the forum.

Especially because the fix is easy. Just post a new thread with a link to the old one.

I know that some (or maybe all) of these issues can be addressed with different forum software. I really like Discourse, although it suffers from some of the same issues. And I know there are exceptions to all of the above. But if you're asking why forums (specifically the power users on those forums) frown on bumping old threads, these are some of the reasons why.

[retina]

Originally Posted by Batalov

You contradict yourself - right there.

"New topic or old topic, " ... the damage has already been done! You (as a user) go in to read some interesting discussions and at the top you see threads at the top:

• Record probable prime found!
  
• The probable primes in Five-or-Bust
  
• Record 73-digit factor is found by user...

and open them - only to find that they were -posted to float up.

Also I have a distinct feeling that you didn't read what I wrote entirely. - https://communitybuilding.stackexcha...active-threads
(see the link for the full discussion)
There you would have found:

Your appeal to authority aside, I don't see how a spammer is related to the new/old threads. Entirely separate issue. And a lot of that dissertation you copied here seems to assume a post into an old topic automatically won't be relevant or be , but we have intelligent users here and adding new information can be useful on occasion.

Someone can start a new topic with also, how does that suddenly make it better in a new topic? People still have to open it to judge it's worth, just the same as with an "old" topic. I don't see what problem is being solved by this.

[Batalov]

New threads started by a new user are, of course, preferred. You are immediately seeing what you will be getting by reading the thread -- before reading the thread.

Our time is free, but it's not worthless.

[paulunderwood]

Originally Posted by bbb120

2887148238050771212671429597130393991977609459279722700926516024197432\
3037991527331163289831446392259419778031109293496555784189494417409338\
0561511397999942154241693397290542371100275104208013496673175515285922\
6962916775325475044445856101949404200039904432116776619949629539250452\
6987193290703735640322737012784538991261203092448414947289768854060249\
76768122077071687938121709811322297802059565867
this 397 digits composite number pass the miller rabin test base from 2 to 306，
so you should use at least one lucas test on probable prime !

Interesting. Many packages rely solely on M-R, e.g. GMP and can be so fooled. With a small cluster it is possible to construct such composites and break cryptosystems. The wise implementers use a Lucas PRP test too.

[Dr Sardonicus]

Originally Posted by paulunderwood

Interesting. Many packages rely solely on M-R, e.g. GMP and can be so fooled. With a small cluster it is possible to construct such composites and break cryptosystems. The wise implementers use a Lucas PRP test too.

The January 1995 paper RABIN-MILLER PRIMALITY TEST: COMPOSITE NUMBERS WHICH PASS IT gives a 337-digit composite which "passes" RM for all prime bases less than 200. It subsequently says

The only limitation towards finding strong pseudoprimes to more bases in this way seems to be the difficulty of doing computations involving such large numbers.

So finding a number that "fools" RM for prime bases up to 300 is not terribly surprising.

IIRC there's a paper showing that a quadratic PRP test delivers something like nine times the bang as the RM test, for four times the bucks.

I checked the 397-digit number n in Pari-GP; first, ispseudoprime(n) gave 0 (composite). Then, ispseudoprime(n, 1) [RM with 1 base] gave 1 (PRP), but ispseudoprime(n,2) [RM tests with 2 "randomly selected" bases] gave 0 (composite).

[paulunderwood]

There is a slight computational overhead for using larger bases (b) in M-R, since in left-right exponentiation we repeatedly calculate:

s*s mod n
s*b mod n if bit is 1

[kriesel]

Originally Posted by retina

Why is it necessary to start a new topic? I haven't seen anything on this board that disallows responding to an existing topic with pertinent information.

Amen to that. I find it useful to have the history of, say CUDALucas development, or gpuowl, nearly all in one thread each, daunting though it may be to read through it all. Launching a new thread at the drop of a hat, scattered all over the haystack of thousands of threads, makes getting an overall picture of a given application more challenging, and can create an annoying amount of clutter.

Granted, there are other considerations, per uncwilly for example.
There may be about as many perspectives on some of the possible tradeoffs as there are participants.

[R. Gerbicz]

Originally Posted by paulunderwood

There is a slight computational overhead for using larger bases (b) in M-R, since in left-right exponentiation we repeatedly calculate:

s*s mod n
s*b mod n if bit is 1

You can almost hide that cost, see for example: https://en.wikipedia.org/wiki/Expone..._window_method
The fast gmp is also using this technique.

[paulunderwood]

Originally Posted by R. Gerbicz

You can almost hide that cost, see for example: https://en.wikipedia.org/wiki/Expone..._window_method
The fast gmp is also using this technique.

I recall this method... for example: N=(1001111011)_2 can be written as (10[10000-1][100-1])_2. The former takes 9S + 6M and the latter takes 9S + 2M +2D (where S=squaring, M=mult by base and D=div by base). I don't think I have it quite right